documentos de trabajo - central bank of chile...ing and subject, but now we are able to distinguish...
TRANSCRIPT
DOCUMENTOS DE TRABAJODoes the Exposure to the Business Cycle Improve Consumer Perceptions for Forecasting? Microdata Evidence from Chile
Fernando FaureCarlos A. Medel
N° 888 Septiembre 2020BANCO CENTRAL DE CHILE
BANCO CENTRAL DE CHILE
CENTRAL BANK OF CHILE
La serie Documentos de Trabajo es una publicación del Banco Central de Chile que divulga los trabajos de investigación económica realizados por profesionales de esta institución o encargados por ella a terceros. El objetivo de la serie es aportar al debate temas relevantes y presentar nuevos enfoques en el análisis de los mismos. La difusión de los Documentos de Trabajo sólo intenta facilitar el intercambio de ideas y dar a conocer investigaciones, con carácter preliminar, para su discusión y comentarios.
La publicación de los Documentos de Trabajo no está sujeta a la aprobación previa de los miembros del Consejo del Banco Central de Chile. Tanto el contenido de los Documentos de Trabajo como también los análisis y conclusiones que de ellos se deriven, son de exclusiva responsabilidad de su o sus autores y no reflejan necesariamente la opinión del Banco Central de Chile o de sus Consejeros.
The Working Papers series of the Central Bank of Chile disseminates economic research conducted by Central Bank staff or third parties under the sponsorship of the Bank. The purpose of the series is to contribute to the discussion of relevant issues and develop new analytical or empirical approaches in their analyses. The only aim of the Working Papers is to disseminate preliminary research for its discussion and comments.
Publication of Working Papers is not subject to previous approval by the members of the Board of the Central Bank. The views and conclusions presented in the papers are exclusively those of the author(s) and do not necessarily reflect the position of the Central Bank of Chile or of the Board members.
Documentos de Trabajo del Banco Central de ChileWorking Papers of the Central Bank of Chile
Agustinas 1180, Santiago, ChileTeléfono: (56-2) 3882475; Fax: (56-2) 3882231
Documento de Trabajo
N° 888
Working Paper
N° 888
Does the Exposure to the Business Cycle Improve Consumer
Perceptions for Forecasting? Microdata Evidence from Chile
Abstract
Given the lag with which the official consumption series are released and the anticipation with which
consumption confidence indicators are published, there is latent attention in assessing the predictive
power of consumer confidence over consumer spending. In this article, we make use of a rich and
unique individual-level dataset obtained by merging a consumer sentiment survey with an
unemployment survey of the Greater Metropolitan area of Chile’s capital city. This is aiming to: (i)
analyse how forecasts of private consumption and components are improved by consumer sentiment
indexes constructed using the labour market characteristics of respondents and re-focusing aggregate
answers according to timing and subject, and (ii) analyse to what extent a greater exposure of
consumers/workers to the business cycle comes out as a better capability to nowcast and short-term
forecast consumption. Considering that consumers work in sectors where decisions are taken based
to a greater or lesser extent on the current and future economic situation, we assume a learning
mechanism permeable to the sentiment of workers that is present when they are surveyed as
consumers, referred to as “exposure”. Since the exposure is extensible to investment decisions, we
include total investment and its components in the analysis. By means of auto-regressions, our
results suggest that the in-sample adjustment of a simple model improves in most cases when any of
the newly built factors are included. Although there is no silver-bullet definition of the factor helping
all aggregates across all horizons simultaneously, future-oriented factors as well as information
coming from consumers working together in Commerce, Industry, Construction, and Personal
Services sectors provide material predictive gains, ranging from 4.5% to 40.2% at one-year
compared to the non-augmented auto-regression. Although surveys of consumers’ perceptions
provide rich information, these results are important since it is necessary to disentangle ideally from
the most disaggregate level to have a better revealing economic behaviour.
Resumen
Dado el rezago con el cual se publican las series oficiales de consumo y la anticipación con la que se
publican los indicadores de confianza de los consumidores, existe una latente atención en evaluar el
poder predictivo de la confianza de los consumidores sobre el gasto en consumo. En este artículo,
utilizamos una rica y única base de datos a nivel individual obtenida al fusionar una encuesta de
sentimiento de los consumidores con una encuesta de desempleo de la Gran Área Metropolitana de la
capital de Chile. Esto tiene como objetivo: (i) analizar cómo mejoran las proyecciones de consumo
privado y sus componentes mediante índices de sentimiento de los consumidores construidos
We thank comments and suggestions of José Miguel Alvarado, Pilar Cruz, Camila Gómez, Carlos Madeira,
Pablo Medel, Michael Pedersen, and Pablo Pincheira. Also, we thank Consuelo Edwards for editing services.
Any error or omission that remains at our own responsibility. The views and ideas expressed in this paper do
not necessarily represent those of the Central Bank of Chile or its authorities. E-mails: [email protected];
Fernando Faure
Universidad de Chile
Carlos A. Medel
Central Bank of Chile
utilizando las características del mercado laboral de los encuestados, y reenfocando las respuestas
agregadas según su temporalidad y sujeto, y (ii) analizar en qué medida una mayor exposición de los
consumidores/trabajadores al ciclo económico se manifiesta como una mejor capacidad para
pronosticar el consumo a corto plazo actual y a futuro. Considerando que los consumidores trabajan
en sectores donde se toman decisiones, en mayor o menor medida, basándose en la situación
económica actual y futura, asumimos un mecanismo de aprendizaje permeable al sentimiento de los
trabajadores que se presenta cuando se les encuesta como consumidores, denominado “exposición”.
Dado que la exposición es extensible a las decisiones de inversión, incluimos la inversión total y sus
componentes en el análisis. Mediante autoregresiones, nuestros resultados sugieren que el ajuste
dentro muestra de un modelo simple mejora en la mayoría de los casos cuando se incluye alguno de
los factores construidos. Aunque no existe una un factor tipo bala de plata que ayude a todos los
agregados en todos los horizontes simultáneamente, los factores reorientados hacia el futuro, así
como la información que proviene de los consumidores que trabajan juntos en los sectores de
Comercio, Industria, Construcción y Servicios Personales, entregan ganancias predictivas
importantes, que van desde 4.5% a 40.2% a un año en comparación con la misma autoregresión sin
aumentar. Si bien las encuestas sobre las percepciones de los consumidores entregan una rica
información, estos resultados son importantes ya que sugieren que necesario indagar, idealmente, al
nivel más desagregado para obtener un comportamiento económico más revelador.
1 Motivation, objectives, and main results
The confidence of consumers has played a key role in the development of modern theoretical economics. More-over, plenty of empirical literature has advocated analysing the relationship between consumer confidence andconsumer spending across world economies. This is relevant for different economic agents having to take impor-tant and irreversible decisions, and policymakers interested in future movements that shape the typically biggestcomponent of domestic demand. Given the lag with which the official consumption series are released, and theanticipation with which consumption confidence is published, there is latent attention in assessing the predictivepower of consumer confidence numeric indicators over consumer spending at the aggregate level.
This paper goes one step further by using microdata evidence of consumer sentiment predictability over aggregateconsumption and investment, taking advantage of a specific survey on consumers’ perceptions which, at the sametime, is linked to a survey of employment/unemployment of the Greater Santiago, Chile’s capital city. These arethe Survey of Employment and Unemployment in Greater Santiago (Encuesta de Ocupación y Desocupación en elGran Santiago, henceforth labelled as "EOD") and the Survey of Perception and Expectations on the EconomicSituation in Greater Santiago (Encuesta de Percepción y Expectativas sobre la Situación Económica en el Gran Santiago,"IEE"), both produced by the Centro de Microdatos of Universidad de Chile. The key advantage is that it is possibleto link, with no error, the perception of a consumer with her/his characteristics as a participant of the labourmarket, e.g. labour situation, economic activity, gender, income, and location. Thus, it is possible to exploitdatabase’s richness by re-grouping consumer perceptions indexes based on labour-market characteristics, aimingto improve their forecasting ability over macro aggregates and hypothesis testing. This treatment also aims todisentangle labour market characteristics more akin to the forecasting task.
By using the mentioned dataset, the aim of this article is two-fold: (i) to analyse the extent to which consumersentiment indexes better forecast aggregate consumption and its components, by constructing new indexes usingthe labour market characteristics of respondents and re-focusing aggregate answers according to timing and sub-ject, and (ii) analyse to what extent a greater exposure of consumers/workers to the business cycle comes out witha better capability to nowcast and short-term forecast consumption, compared to consumers/workers that partic-ipate in economic sectors less exposed to the business cycle. This latter exercise seeks to find the most accurateinformation of the survey when tracking the economy. Our claim assumes that consumers work in sectors wheredecisions are taken based to a greater or lesser extent on the current and future economic scenario. The consumerscan take these decisions themselves, be aware of them, or be affected by their consequences. Thus, we assume theexistence of a learning mechanism permeable to the sentiment of workers that is present when they are surveyedas consumers. We call this mechanism as "exposure" and we operationalize it by means of the correlation of work-ers’ economic sector with the output gap. Since this argument is extensible to investment decisions, we includetotal investment and its components in the analysis.
After checking the in-sample consumer-survey-based statistical inference within the simplest (and more trans-parent) econometric setup, we test our two proposals by means of out-of-sample comparisons exercises.1 Thus,the first hypothesis comes out as a multi-horizon forecast comparison between an atheoretical predictive autore-gressive model for a given macro aggregate and its corresponding survey-based factor-augmentation version,using different computations of consumer indexes. The second hypothesis consists of an out-of-sample compar-ison between two survey-based factors, one grouping indicators from consumers currently working in sectorsmore sensitive to the business cycle, and the other conformed with consumers working in the remaining sectorssurveyed. We proxy the business cycle by means of an output gap estimation.
Note that these hypotheses are tested for Private consumption, Non-durable consumption, and Durable con-sumption, plus Gross fixed capital formation (GFCF), Construction & works, and Machinery & equipment. Wealso make use of the same re-classification of answers used in Chanut, Marcel, and Medel (2019) in terms of tim-ing and subject, but now we are able to distinguish the economic sector where the consumer is currently working,adding new dimensions to exploit.
1These exercises consider a fixed vintage of the macroeconomic aggregates, which are subject to a revision policy. Thus, exercises are notconducted with real-time data.
1
Our in-sample results suggest that the adjusted goodness-of-fit coefficient of the first-best auto-regression im-proves in the majority of cases when including any of the considered consumer-perception-based factors. Ourbaseline results show that using specifically the re-classification by subject and timing, the Personal Future Expec-tation Index as well as the Overall Future Expectation Index improve for almost all macro aggregates consideringany sectorial definition of the factor. In contrast, the Personal Current Situation Index is less helpful to explainthe dynamics of any macro aggregate built with any sector. Most of these improved adjustments are found forGFCF and its components, and for Durable consumption to a lesser extent. Consumers working in Commerce andIndustry provide more useful information than remaining sectors, as well as when adding consumers in Construc-tion and Personal Services sectors without manufacturing. When accounting for the statistical significance of theconsumer-perceptions factor, there is a concentration of significant cases using future-based indexes for the samemacro aggregates, which exhibits better adjustments. Also, the evidence is mostly favourable for sectors moreclosely related to the business cycle rather than those that are less so or plainly insensitive to it.
Our out-of-sample results show, with a baseline added to alternative setups, predictive gains at all horizons for allvariables, in some cases of a sizable magnitude whereas in others non-statistically significant. The most fruitfulresults are obtained for Non-durable consumption and GFCF, which are statistically superior to the predictivebenchmark model when made up by sectors more akin to the business cycle. There is no silver-bullet definitionof the factor helping at the same time all aggregates across all horizons. However, in more consumption seriesthan not, the Personal Country Situation Index and the Country Current Situation Index lead to the best results. Inturn, the Overall Future Expectation Index does so for investment series. Regarding the economic sectors in whichconsumers participate and provide more useful predictive information, unquestionably the best forecasting resultsare concentrated in Commerce, Industry, Construction, and Personal Services (thus, out of 3,060 individuals surveyedper quarter, this factor is composed of 1,735 answers per quarter on average) and in Commerce, Construction, andPersonal Services (comprising 1,313 answers per quarter on average). When comparing between factors, the resultsare largely favourable to the factors based on consumers more exposed to the business cycle. In this line, themost fruitful results are obtained with Commerce, Industry, and Transportation sectors and Commerce, Construction,and Personal Services that outperform the remaining sectors less exposed to the business cycle. Although surveysof consumers perceptions provide rich information, these results are important as it is necessary to disentangleideally from the most disaggregate level to have a better revealing economic behaviour.
The remainder of the article proceeds as follows: in section II, we review the related literature on this topic. Insection III, we describe our econometric setup in terms of the individual and macro aggregates data, detailing theinformation contained in the two surveys and the construction of all consumer-based factors. It also describes thestatistical inference carried out to both in- and out-of-sample analyses. In section IV we display our results of thechosen baseline autoregressive model, and the in-sample and predictive evidence. Finally, section V provides theconclusions drawn from the study.
2 Literature review
Several articles set up the baseline stylised facts and common findings on this matter, specially analysing the caseof advanced economies such as the Euro Area and the United States; for the latter, specially using the Survey ofConsumers (SoC) produced by the University of Michigan.2 For instance, Fuhrer (1993) studied whether the SoCcan explain different economic phenomena with simple predictive models and a vector auto-regression (VAR).Their results suggest that a good part of the variation in the sentiment index is explained by macro aggregates,and that the SoC explains and improves a little on the actual consumption series and its forecasts. Carroll, Fuhrer,and Wilcox (1994) analyse if the information of the SoC predicts changes in private consumption, finding thatwithin a reduced-form model, the SoC anticipates consumption growth. Howrey (2001) makes use of part of theSoC to analyse the ability to anticipate recessions and recovery episodes using a VAR, finding that the SoC isbarely superior to a benchmark model when forecasting GDP one step ahead. Doms and Morin (2004) mimicsSoC answers using newspapers-based indexes, finding three transmission channels to consumers perceptionsthrough actual economic data, tone, and volume. The newly-built indexes come out as playing a role whenconsumers update their perceptions, which is directly dependent on the volume of economic news. Ludvigson
2For more details on the full content of the SoC, see Survey of Consumers, University of Michigan (2020).
2
(2004) shows that the SoC contains valuable information on the future path of aggregate consumption but yieldingmodest predictive gains, whereas the same SoC information could be found in economic/financial indicators.By exploiting the microdata of SoC, Souleles (2004) finds that aggregate shocks do not impact all householdsequally, and demographics explain this heterogeneity to a large extent. Particularly, people’s forecast errors aresystematically correlated with their demographics, and sentiment helps forecast consumption growth.
Geiger and Scharler (2019) use the SoC to show how consumers assess gasoline price shocks using a sign-restrictedVAR, finding that oil prices and unemployment guide consumer perceptions, and these perceptions are consistentwith economic developments. Lahiri, Monokroussos, and Zhao (2015) analyse the SoCs’ predictive ability overconsumption using a dynamic factor model and a large, real-time, jagged-edge dataset, finding a robust rolefor perceptions in improving consumption-forecast accuracy particularly through the services component. Allthese results are promising and call to analyse its potential extension to other countries with their correspondingsurveys. Moreover, valuing the predictive content of surveys for macro aggregates, and the way how it couldbe better handled, Bürgi and Sinclair (2015) suggests a nonparametric approach to refine the pool of respondentsparticipating in the Survey of Professional Forecasters conducted by the Federal Reserve of Philadelphia. Theirapproach leads to significant short-term predictive gains, which could certainly be interesting to further analysein the context of this article.
This list is not intended to be comprehensive; rather, it aims to show the evolution of the type of questions andresults that a rich consumer sentiment survey could provide for economic analysis. Outside the United States,the evidence is also substantial. Similar exercises to that conducted with SoC has been produced with specificconsumer sentiment and perceptions indexes for the Euro Area (Gayer, 2005; Girardi, Golinelli, and Pappalardo,2017; Basselier, Liedo, and Langenus, 2018), Germany (Schröeder amd Hüfner, 2002; Abberger, 2007), Italy (Brunoand Malgarani, 2002), Sweden (Hansson, Jansson, and Löf, 2003; Österholm, 2014), China (Li, 2011), South Korea(Moon and Lee, 2012), and Garnitz, Lehmann, and Wohlrabe (2017) for 44 countries. De Mello and Figuereido(2014) provide evidence for the Brazilian sentiment indicators acting as covariant when forecasting economicactivity in an econometric setup similar to this article’s.
In emerging market economies, as is the case of Chile, the evidence is scarce. Nonetheless, recent research suggeststhat aggregate consumer sentiment indicators do not perform well when trying to predict private consumptionand investment within a simple, statistical ensemble. However, when re-classifying survey answers according totiming (current/future) and subject (personal/country) dimensions, the evidence turns favourable to the use offorward-looking factors to predict specific components of the aforementioned macro aggregates (Chanut, Marcel,and Medel, 2019). By making use of an instrumental definition of consumer sentiment based on a different con-sumer confidence survey, Acuña, Echeverría, and Pinto-Gutiérrez (2020) find that consumer sentiment improvesin-sample modelling of private consumption as well as its out-of-sample forecast in a fundamental-based envi-ronment. Their results are robust to nonlinearities and, to some extent, they contradict theoretical predictions ofconsumer spending theories based on, for instance, the Precautionary Saving Hypothesis (Blanchard and Mankiw,1988).
Furthermore, using a subset of questions from a consumer perceptions survey, Albagli et al. (2019) find thatconfidence shocks impact consumption negatively up to eight quarters after returning to its previous growth rate,providing some support to consumer sentiment indicators acting as predictors. Also, Figueroa and Pedersen(2019), by picking specific questions of the same consumer confidence survey used in Chanut, Marcel, and Medel(2019) and Albagli et al. (2019), analyse its forecast ability on Chilean economic activity. Authors’ findings arefavourable to the use of consumer answers for horizons different from nowcast, suggesting that consumers indeedprovide good-quality frontward information.3 Interestingly, all these results have in common at least two limitingcharacteristics: (i) the need for a re-classification or refinement of quantitative indexes in search of actual harddata predictability, and (ii) the use of indexes in an aggregate way, in the sense that the mentioned re-shaping ofindicators makes use of all individuals’ data whatever the driver of the re-classification is, classified by eithertiming or subject.
3Some other applications of consumer sentiment indicators in Chile rely on its use as an aggregate covariate in a data-rich ensemble toforecast activity and inflation. See, for instance, Calvo and Ricaurte (2012) making use of one particular survey question, and Riquelme andRiveros (2018) using disaggregated survey indicators to build a coincident indicator of total monthly economic activity.
3
3 Econometric setup
Firstly, it is relevant to distinguish between consumer expectations and consumer sentiment. In this article, we makeuse of consumer sentiment, which means the economic agents’ opinion on future relevant economic developmentsthat may be influenced by today’s actions (e.g. "Do you think the state of the economy has improved, stayedthe same, or deteriorated over the last year?"), in contrast to economic expectations in which agents state theirperceived most likely value on a particular targeted variable (e.g. "What do you think will be the year-on-yearrate of CPI inflation in the next quarter?"). It is generally believed that qualitative questions to report agents’sentiment, while less accurate, are better understood by a non-expert audience and, thus, better reflect their truesentiment, as opposed to a random guess. This is precisely a feature exploited in this article, testing whetherconsumers diverge in their opinions depending on the economic sector in which they participate. Naturally, thereis no wrong answer to a sentiment survey, whereas expectations allow us to even characterise a numeric errorterm fully.
A review of the econometric exercise conducted in this article is the following. First, we search the best stationaryautoregressive AR(p) forecasting model for each of the six macroeconomic variables considered. We search acrossp lags, p={1,...,4}, as actual macroeconomic data is released in quarterly frequency. Instead of using the traditionalin-sample information criteria to select the model with the best fit, we use a more demanding strategy by choosingthe best out-of-sample model when forecasting from 1 up to 4 quarters ahead. This is also the reason why wesearch among the simplest family of AR(p) models, i.e. to avoid the issue of overfitting by including regressorsthat exaggerate the dynamics of the series at the cost of spoiling out the predictive power of the model. Wealso choose this strategy to stress out the out-of-sample benefits of the exogenous factor without including anyadditional covariate.4
Secondly, we compare the multi-horizon forecast of the chosen AR(p) model extended with the exogenous consumer-survey-based factor. By performing this exercise, we can determine if this extension is fruitful from an out-of-sample perspective–and checking for our first hypothesis. We compare these forecasts by means of the root meanforecast error (RMSFE) and using the Clark and West (2007) test for statistical inference.
Thirdly, we build the exogenous consumer-survey-based factor by grouping consumers’ answers. These con-sumers are currently working in sectors that are more sensitive to the business cycle. This makes us consider ashare of total answers that compound the factor, and the complementary share compounding another factor withconsumers working in other sectors less sensitive to the business cycle. Hence, we compare the influence of bothfactors on the predictive ability of the chosen AR model, expecting that the consumers possibly more aware ofthe business cycle would provide more accurate information on their economic situation–and checking for oursecond hypothesis. This comparison is made by means of the RMSFE and the Harvey, Leybourne, and Newbold(1997) test. On top of this, we also re-classify some answers according to timing and subject. Thus, factors containthree dimensions: timing, subject, and economic sector, e.g. "Personal Current Situation Index of consumers workingin Commerce and Industry". An important drawback of this exercise is that it is not conducted in real time becauseof data vintages availability.
Fourthly, we try four different specifications of the economic sectors when building consumer-survey-based ex-ogenous factors. This is because the sensitivity to the output gap of the economic sectors of the surveyed con-sumers is evolving with the time difference (lags) with which the correlation coefficient is calculated.
Finally, we conduct a robustness exercise using the whole range of AR(p) models available combined with dif-ferent setups of the exogenous factor, namely, first differences or levels, and the number of terms included in theforecasting equation. These exercises are relevant because in some cases the best forecasting model is evolvingwith the forecast horizon for a given macro aggregate, and comparisons must be made with the best benchmarks.Also, there is no single manner to include the factor in the forecasting equation, and so, we try different specifica-tions.
4In Acuña, Echeverría, and Pinto-Gutiérrez (2019), the authors use another consumer-survey aggregate factor to predict private consump-tion with a fundamentals-based consumption model. The authors, however, do not check if the non-augmented model is better than anatheoretical simple benchmark. In this sense, the exercise conducted in this article is more motivated by the forecasting results.
4
3.1 Consumer perceptions, labour market microdata, and the business cycle
As mentioned above, we make use of the microdata freely available (after submitting an online registry) by theUniversidad de Chile’s Centro de Microdatos (http://documentos.microdatos.cl/). This database is the result ofmerging the datasets of two surveys: the Survey of Employment and Unemployment in Greater Santiago (the EOD) andthe Survey of Perceptions and Expectations on the Economic Situation in Greater Santiago (the IEE). A unique feature isthat both databases are available (anonymously) at the individual level and are already merged since respondentsare asked about both their labour situation as a worker and her/his economic perception as a consumer. Theindependence in answering both surveys, especially that of sentiment, is ensured by the wording of the questions.Thus, the sentiment is not conditioned to her/his labour situation by survey design.
Naturally, the universe of the IEE is the same of that of EOD:5 inhabitants over 14 years old living in the San-tiago Metropolitan Region and in Puente Alto and San Bernardo counties. These add up to about 40% of theChilean population in 2017. The sample is composed of 3,060 individuals per quarter, consisting in a stratifiedrandom sampling with a panel data component–a rotating panel. In this method part of the whole panel is keptpermanently, and another part is a completely new cross-section sample. The sample of 3,060 individuals in eachquarter is divided into four subsamples of 765 individuals, where each subsample is independent and representsthe Greater Santiago. The rotation design is of the 2-2-2 type, where two selected individuals are interviewedtwice in a row, then they are not contacted in the following two rounds, and then they are interviewed again inthe following two rounds, covering a period of 18 months in total. The collection technique is face-to-face inter-view and the reported answer rate reaches 77.4% (informed on March 2014). The representativeness with respectto the universe could be considered adequate, but this is not the case when considering the whole country, asit focuses in Santiago only.6 The IEE is released quarterly and fully available since March 2001 (75 observationsavailable up until September 2019).
Note that across all questions asked in IEE there is a time dimension and a subject dimension, allowing us tore-classify them in these terms and the economic sector reported by each individual. These questions are thefollowing, responded in a qualitative manner (with two exceptions):
• Q1: How was the economic situation of the country a year ago? (better, same, worse)
• Q2: What are the three main problems of the country? (free answer)
• Q3: How did the income of your household vary in the last 12 months? (increased, same, decreased)
• Q4: How is the situation of your household in terms of indebtedness? (complicated, average, no problem)
• Q5a: Did a member of your household buy a durable good in the past three months? (yes/no)
• Q5b: If so, how have you financed it? (credit/cash)
• Q6: In one year, how will be the economic situation of the country compared to today? (better, same, worse)
• Q7: How will the income of your household vary within the next 12 months? (increase, same, decrease)
• Q8: What will be the CPI inflation rate in 12 months? (numeric answer)
• Q9a: Will a member of your household buy a durable good in the next three months? (yes/no)
• Q9b: If so, how will it be financed? (credit/cash)
• Q10: Are you or is a member of your household thinking of buying a house in the next 12 months? (yes/no).
5Specific details can be found in Centro de Microdatos (2016).6Santiago is Chile’s capital city with a population of 7.037 million, i.e., 37.57% of the country’s total of 18.730 million as of 2017.
5
It is important to look at the way in which qualitative answers are transformed into quantitative indicators. Thereare two steps. The first consists of aggregating the individual answers to the same question to get a number called"balance statistic," while the second step is to aggregate the balance statistic of each question to form a syntheticindicator. Intuitively, the balance statistic is the difference between the percentage of positive answers and thepercentage of negative answers.7 Any synthetic indicator (step 2) is constructed by taking the equally weightedaverage of the balance statistic of the questions composing the indicator.
Similarly to what Chanut, Marcel, and Medel (2019) sugget, we overcome the limitations of an aggregate indicatorthat blends all questions without a specific focus by re-grouping subsets of questions distinguishing betweentiming (current sentiment/future expectations) and subject (personal and country-wide ). Crossing these twodimensions originates four different indicators, in which the questions are mapped as:
1. Country Current Situation Index (CCSI ): Q1,
2. Country Future Expectation Index (CFEI ): Q6,
3. Personal Current Situation Index (PCSI ): 13Q3 + 1
3Q4 + 13Q5a,
4. Personal Future Expectation Index (PFEI ): 13Q7 + 1
3Q9a + 13Q10,
5. Overall Future Expectation Index (OFEI ): 14Q6 + 1
4Q7 + 14Q9a + 1
4Q10.
Note that we add a fifth indicator (OFEI) overriding the "subject" dimension (mixing country and personal) inorder to have an appraisal on the overall predictive ability of those answers. In turn, EOD asks for traditionalpersonal, demographic, and economic information: county, gender, age, marital status, educational level, em-ployment situation, reason of unemployment, availability, occupational position, number of workers hired underyour responsibility, economic activity, type of institution or establishment, total hours worked during the week ofreference, total days worked during the week of reference, duration of inactivity, willingness to work extra hours,effort in job search, reason for no job searching, type of contract, salary, income from independent activities, pen-sion, and total household income. Breaking over the "Employment Situation" answers (whose answers could beWorking, Unemployed, Studying, Housework, and Other), and particularly mapping the "Working" answer with the"Economic Activity", we found that 55.2% of the sample that reports working (average 2001.I-2019.III), 22.5% dothat in the Commerce sector, 18.2% in Commonality and Social Services, 15.6% in Public and Financial Services, 13.8%in Industry, 8.0% in Construction, 7.8% in Transportation, and the rest in four remaining smaller sectors (see Figure1).
Thus, our hypothesis propose that consumers in economic sectors that are more exposed to the business cycle couldbe able to forecast better overall consumption and investment than (i) a reasonable simple benchmark, and (ii)consumers in sectors less sensitive to the business cycle. But which is the group of consumers more exposed to thebusiness cycle? In our setup, this is equivalent to asking which economic sectors are more correlated to (a measureof) the output gap. There is no single way on how to determine the sensitivity or exposure of specific economicsectors to the business cycle. For instance, it is possible that those sectors with a higher GDP weight could bethe most exposed, but it ignores the dynamic relationship with the output gap. Thus, it could be the case that asector, while having a heavy weight on GDP, fluctuates less than other sectors with lighter weight throughout thebusiness cycle. This is actually the case of Financial Services (14.9%) and Mining (9.4%) with a weight greater than,for instance, Transportation (5.0%), but this latter with a greater correlation with the (lagged) output gap. Also, thecorrelation is calculated using the output gap instead of GDP to differentiate between, for instance, episodes of"GDP growth" from "GDP growth above the potential GDP".
7Formally, it is defined as balance statistic=100(∑Ni=1 ωixi), where N is the total number of individual surveyed, ωi is the weight of the
i-th sample unit, and xi is the response of the i-th sample unit, taking value one when the answer is "yes" or "increase", minus one when theanswer is "no" or "decrease", and zero when the answer is "stable" or "same". The weight of the i-th sample unit is chosen in such a way thatthe sample is representative: the weight is inversely proportional to the probability of unit i to be drawn; hence, if the sample is random, theweight is simply ωi=1/N.
6
Figure 1. Composition of the Centro de Microdatos’ employment situation answersby situation and economic activity, 2001-2019 average (*)
55.2%
5.9%5.1%
16.8%
17.0%
Working UnemployedStudying Housew orkOther
I. Employment Situation
13.8%
8.0%22.5%
15.6%
12.4% 18.2%
7.8%
Agriculture MiningIndustry ConstructionCommerce Pub. & Fin. S.Personal S. Comm. S.Transp. S. Other
II. Employment by Economic Activity
(*) Source: Authors’ calculations based on Centro de Microdatos database.
The approximation used in this article for the output gap is obtained as the difference between the logarithmiclevel of actual GDP and potential GDP.8 The latter is defined as the logarithmic level of the seasonally-adjustedand filtered version of the GDP including up to five years of forecast observations coming from an ad-hoc ARIMAmodel. This last step is performed to avoid the "end-of-sample" identification problem when using the Hodrick-Prescott (λ=1,600) method to filter the series. The seasonal adjustment program used is the X-13ARIMA-SEATS,whereas the ARIMA forecasting model is the so-called airline model (Box and Jenkins, 1970; Ghysels, Osborn, andRodrigues, 2006).
Figure 2. Output gap and selected activity sectors (annual variation) (*)
20
16
12
8
4
0
4
8
12
16
20
96 98 00 02 04 06 08 10 12 14 16 18
Output gap Construction Commerce Personal S.Industry Financial S. Public Adm. MiningTransportation
Per
cent
Percent
(*) Source: Authors’ calculations based on Central Bank of Chile’s database.
8The source of the GDP series is the Central Bank of Chile’s Quarterly National Accounts database(https://si3.bcentral.cl/Siete/secure/cuadros/home.aspx). We use the vintage available up to 2019.III.
7
We identify the economic sectors most sensitive to the output gap by means of cross correlation calculated withthe current and up to 4-quarter lagged output gap. Particularly, the correlation in period t of the activity sectorsConstruction, Commerce, Personal Services, Industry, Financial Services, Public Administration, Mining, and Transporta-tion, obtained from the Central Bank of Chile’s Quarterly National Accounts database, with the correlation in t, t-1,t-2, and t-3 quarters of the output gap. The time series plot of involved variables is presented in Figure 2, and theestimated correlations are presented in Table 1.
Table 1: Correlation coefficient between (lagged) output gap and economic activity sectors (*)Output gapt Output gapt−1 Output gapt−2 Output gapt−3 Output gapt−4
Output gapt 1.000 0.918 0.791 0.634 0.495Output gapt−1 0.918 1.000 0.918 0.789 0.636Output gapt−2 0.791 0.918 1.000 0.917 0.790Output gapt−3 0.634 0.789 0.917 1.000 0.921Output gapt−4 0.495 0.636 0.790 0.921 1.000Miningt 0.105 0.031 -0.013 -0.036 -0.040Industryt 0.330 0.108 -0.151 -0.362 -0.533Constructiont 0.541 0.477 0.302 0.068 -0.137Commercet 0.507 0.303 0.057 -0.190 -0.378Transportationt 0.213 0.072 -0.111 -0.315 -0.486Financial Serv.t 0.262 0.183 0.052 -0.126 -0.301Personal Serv.t 0.438 0.347 0.198 0.072 -0.058Public Adm.t 0.187 0.213 0.257 0.327 0.430
(*) Shaded cells=bigger correlation coefficient (≥ |30%|) for a comparable time difference.Sample: 2001.I-2019.III (75 observations). Source: Authors’ calculations.
Note that, as the sectors with an acceptably high correlation with the output gap are evolving with the lag withwhich the correlation is calculated, we make use of four different definitions of the factor combining more infor-mative sectors, thus:
1. CI: Commerce + Industry
2. CIT: Commerce + Industry + Transportation
3. CICS: Commerce + Industry + Construction + Personal Services
4. CCS: Commerce + Construction + Personal Services.
Notice that we consider for factor building all those sectors having more than twice a correlation coefficient above|30%| between t and t-4 output gap lags. Also, we opt to exclude Public Administration because it is merged withFinancial Services in the EOD/IEE that, in turn, exhibit a low correlation with the output gap for considered timedifferences. Recall that for our second hypothesis, we compare the predictive ability of each of these four factorsseparately in our AR(p) setup with those composed with the remaining activity sectors surveyed by IEE (seeFigure 1).
3.2 Macroeconomic aggregate data
We perform the same econometric exercise when predicting the six macroeconomic variables with the four con-sumer perception factors. Mentioned variables are Private consumption (PC) plus two out of three of its com-ponents separately—Durable consumption (DC) and Non-durable consumption (NDC)–, and Gross fixed capitalformation (GFCF) plus its two components–Construction & works (CWK) and Machinery & equipment (MEQ).
The source of all these variables is the Central Bank of Chile’s Quarterly National Accounts database; specifically, thevintage available until 2019.III. The transformation used to achieve stationarity is the annual percentage change.The series achieve stationarity with the full sample (2001.I-2019.III, 75 quarterly observations) according to theAugmented Dickey-Fuller (null hypothesis: the series has a unit root) using a trend, a constant, and 4 lags for PC(p-value: 0.193), a trend, a constant, and 4 lags for NDC (p-value: 0.147), a constant and 4 lags for DC (p-value:
8
0.145), a trend, a constant, and 4 lags for GFCF (p-value: 0.192), a constant and 4 lags for CWK (p-value: 0.127),and a trend, a constant, and 8 lags forMEQ (p-value: 0.381). No structural change is found in any of the series.
These macroeconomic aggregates are depicted in Figure 3, panels I and II, along with the indexes depicted bytiming/subject (CCSI, CFEI, PCSI, PFEI, and OFEI) in panels III to VII. Note that all these factors achieve station-arity with a trend, a constant, and 5 lags (but excluding the trend with PCSI and PFEI indexes) according to theAugmented Dickey-Fuller test at the 10% confidence level. This is important because although the factors aredeveloped in levels, they already refer to temporal comparisons according to their wording (see subsection 3.1).So, it could be unusual for one of these indicators to persistently grow over an extended time span. Thus, there isnot an econometric issue when including them in the AR model in levels.
3.3 In-sample and forecast evaluation framework
We use the estimation sample covering from 2001.I to 2014.IV (56 observations) to estimate the AR(1) to AR(4)models for the six macroeconomic variables to determine which of them is the best forecasting model makinguse of the remaining evaluation sample covering from 2015.I to 2019.III (19 observations). After finding the bestmodel, we then include four terms of the consumer-based factor and compare the resulting forecast with that ofthe non-augmented model. In our robustness exercises, we perform different setups on factor specification (levelor first differences) combined with different number of terms included in the auto-regression (from one to four),under different AR(p) models. There is not a clear way on how to include the factor. However, as mentionedabove, consumer data is evolving according to a 2-2-2 rotating panel scheme, and thus, include at least more thanone term imply consider different and ultimately more consumers in the forecasting task. We opt to set the 4-termfirst differences of the factor as a baseline setup, which will be naturally challenged by robustness exercises. Thebaseline specification thus is:
yt = y+P
∑i=1
ρiyt−i + φ1∆ ft + φ2∆ ft−1 + φ3∆ ft−2 + φ4∆ ft−3 + εt, (1)
where yt is any of the macro aggregates, ∆ ft corresponds to first difference of the consumer-based factor andits complementary versions, {y,ρi,φi} are to-be-estimated parameters once P ∈ {1, ..., 4} is determined accordingto its out-of-sample performance without including a factor, and εt is a white noise. Note that the factor entersthe equation with four terms, including a contemporaneous term. This is so because the survey data becomesavailable four to five months prior to macroeconomic aggregates.
For the same estimation sample, we compute the incremental adjusted goodness-of-fit coefficient with and withoutincluding the factor, and the joint statistical significance of the φi-coefficients in equation (1). We choose the Pcoefficient of equation (1) and also conduct the whole forecasting exercise making use of the evaluation sample in arolling window scheme. This is also required for the statistical inference framework.
The forecast evaluation statistic used is the root mean squared forecast error (RMSFE) defined as:
RMSFEh =
[1
P(h)
T+1−h
∑t=R
(yt+h − yt+h|t
)2] 1
2
, (2)
where yt+h|t represents the forecast of yt+h made with information known up until time t. We generate a total ofP(h) forecasts, satisfying P(h)=T + 2− h+ R, where h is the forecast horizon, h={1,2,3,4}, and R is the number ofobservations in the estimation sample (56 observations).
For simplicity, the results are reported using a relative measure of the RMSFE, easing a comparison across thealternative forecasts. Two comparisons are made:
(i) : Relative RMSFE1h=RMSFEFactor augmentedh /RMSFENo factorh , (3)
(ii) : Relative RMSFE2h=RMSFEFactor augmented (more exposed)h /RMSFEFactor augmented (less exposed)h ,
9
Figure 3. Macroeconomic aggregates and consumer-perceptions-based indexes (*)
6
420
246
8
101214
40
302010
01020
30
405060
2002 2004 2006 2008 2010 2012 2014 2016 2018
Private Cons. Durable Cons. Nondurable Cons. (RHS)
I. Macroeconomic Aggregates: Consumption
Per
cent
age P
ercentage
Mean=4.74%; SD=3.15%Mean=10.34%; SD=13.58%Mean=4.04%; SD=2.75%
40
30
20
10
0
10
20
30
40
50
2002 2004 2006 2008 2010 2012 2014 2016 2018
Gross Fixed Cap. Fmtn. Const. and Works Machinery and Equipment
II. Macroeconomic Aggregates: Investment
Per
cent
age P
ercentage
Mean=6.28%; SD=9.83%Mean=4.50%; SD=6.38%Mean=10.14%; SD=19.01%
24
28
32
36
40
44
48
52
56
60
2002 2004 2006 2008 2010 2012 2014 2016 2018
Comm. + Ind. Comm. + Ind. + Trans.Comm. + Ind. + Const. + Pers. Serv. Comm. + Const. + Pers. Services
III. Country Current Situation Index
Inde
x Index
202530354045505560657075
2002 2004 2006 2008 2010 2012 2014 2016 2018
Comm. + Ind. Comm. + Ind. + Trans.Comm. + Ind. + Const. + Pers. Services Comm. + Const. + Pers. Services
IV. Country Future Expectations Index
Inde
x Index
30
32
34
36
38
40
42
44
46
48
2002 2004 2006 2008 2010 2012 2014 2016 2018
Comm. + Ind. Comm. + Ind. + Trans.Comm. + Ind. + Const. + Pers. Services Comm. + Const. + Pers. Services
V. Personal Current Situation Index
Inde
x Index
22
24
2628
3032
34
363840
42
2002 2004 2006 2008 2010 2012 2014 2016 2018
Comm. + Ind. Comm. + Ind. + Trans.Comm. + Ind. + Const. + Pers. Services Comm. + Const. + Pers. Services
VI. Personal Future Expectations Index
Inde
x Index
24
28
32
36
40
44
48
52
56
2002 2004 2006 2008 2010 2012 2014 2016 2018
Comm. + Ind. Comm. + Ind. + Trans.Comm. + Ind. + Const. + Pers. Services Comm. + Const. + Pers. Services
VII. Overall Future Expectations Index
Inde
x Index
(*) Horizontal line in III-VII=mean of the mean of each depicted series. Source: Authors’ calculationsbased on Central Bank of Chile and Centro de Microdatos database.
10
where "more/less exposed" refers to the relationship with the output gap. Thus, values below unity imply a betterperformance in favour of our two proposed hypotheses.
For the first case, i.e. the forecast including the factor compared to the AR version with no augmentation, the Clarkand West (2007; CW) test is used because it consists of an adjusted RMSFE comparison due to model encompass-ing. Hence, the CW test evaluates model adequacy instead of forecast accuracy, as opposed to the Harvey, Leybourne,and Newbold (1997) test used to compare forecast accuracy between factors. Note that the CW is not the mostappropriate for small sample environments due to a reduction of its power. However, there is no small-samplecorrection available and results are presented for reference. Thus, we use it for want of a better alternative.
The CW test can be considered as both an encompassing test and an adjusted comparison of the MSFE. Theadjustment is made to ensure a fair comparison between nested models. Intuitively, the CW test removes a termthat introduces noise when a parameter, that should be zero under the null hypothesis of equal MSFE, is estimated.
The core statistic of the CW test is constructed as follow:
zt+h = (ε1,t+h)2 −
[(ε2,t+h)
2 −(
y1,t+h|t − y2,t+h|t)2]
, (4)
where
ε1,t+h = yt+h − y1,t+h|t, (5)ε2,t+h = yt+h − y2,t+h|t,
represent the corresponding forecast errors. Note that y1,t+h|t and y2,t+h|t denote the h-step-ahead forecasts gen-erated from the two models under consideration. "Model 1" is the parsimonious or small model without factor-augmentation that is nested in the larger "Model 2". In other words, Model 2 would become Model 1 of some ofits parameters were set to zero.
With a little algebra, it is straightforward to show that it could also be expressed as follows:
Sample MSFE-Adjusted=2
P(h)
T+1−h
∑t=R
ε1,t+h (ε1,t+h − ε2,t+h) . (6)
This statistic is used to test the following null hypothesis, against the alternative:
H0 : E [Sample MSFE-Adjusted] = 0, (7)H1 : E [Sample MSFE-Adjusted] > 0.
The CW test suggests a one-sided test for a t-type statistic based upon the core statistic in equation (4), i.e. asymp-totically normal critical values. In most of the analysis, it follows Clark and McCracken (2001, 2005). Theoreticalresults in those papers require models to be estimated with non-linear least squares (which we use in this article)and multi-step forecasts to be made with the direct method. As we make use of the iterated forecast method, weshow the results at more than one-step-ahead horizon just for reference.
For the second hypothesis considered in equation (3), i.e. comparing between the two factor-augmented forecasts,the equal predictive ability of Harvey, Leybourne, and Newbold (1997, HLN) test is used. This consists in a small-sample-corrected version of the Diebold and Mariano (1995) test, and plainly tests forecast accuracy. Accordingto the Diebold-Mariano original test, we focus on testing the following null hypothesis against the alternative:
H0 : E[dt(h)
]= 0, (8)
H1 : E[dt(h)
]6= 0,
wheredt(h) =
(yt+h − y1,t+h|t
)2−(
yt+h − y2,t+h|t)2
. (9)
11
Our null hypothesis posits that forecast generated from the nested model performs equally to those generatedfrom the larger model. As noted by HLN, using an approximately unbiased estimator of the mean’s varianceleads to a modified Diebold-Mariano test statistic:
HLNh =
[n+ 1− 2h+ n−1h(h− 1)
n
] 12 (
dt(h)/σdt(h)
), (10)
which must be contrasted with critical values from a Student’s t distribution with (n-1) degrees of freedom, anddt(h) corresponds to the sample mean and σdt(h) to the standard deviation of the term in equation (9). Notice thatthis test requires that the estimates be made with a rolling window scheme. Thus, we perform our estimations withrolling windows of a fixed size of 50 observations. It is important to emphasize that the two tests differ in a numberof aspects, the most important difference being that they are designed for different purposes. Consequently, weexpect these two tests to deliver different results.
4 Results
As mentioned above, we choose the best AR(p) model following its predictive performance. In Figure 4, we showthe (absolute) RMSFE of the AR(1) to AR(4) models for each horizon and macro aggregate. Two observationsare worth mentioning: (i) there are heterogeneous results across horizons for a given variable, and in some casesthe best model is evolving with the horizon, leading to further analyse other setups in an extensive robustnessexercise (subsection 4.3), and (ii) there are some variables that are more difficult to predict as they exhibit morevolatility than others, shedding some light on where is more likely to have room for improvement. Except for DCat h>1 andMEQ for h={2,3}, the best forecasting model is the AR(1). These two cases also motivate our robustnessexercises, where we analyse the same results but using an AR(4) model. Thus, we make use of this simple—butdemanding in terms of predictability—model to conduct our baseline forecasting exercise, but subject to furtheranalysis.
Figure 4: Multi-horizon RMSFE of the AR(p) models by macroeconomic aggregate (*)
0
1
2
3
4
5
6
7
8
9
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
p=1 p=2 p=3 p=4
RM
SFE
RM
SFE
Horizon:
Private Durable Nondurable Gross Fixed Construction MachineryConsumption Consumption Consumption Capital Fmtn. and Public W. and Equipment
(*) Each bar displays the out-of-sample RMSFE by horizon and macroeconomic aggregate,for a given p-lag, p={1,...,4}, of the AR(p) model. Source: Authors’ calculations.
12
4.1 In-sample diagnostics
Our first in-sample results comprise the incremental adjusted goodness-of-fit coefficient9 when comparing theAR(1) with and without including the consumer-based factor. The results are presented in Table 2, showing theratio between the adjusted goodness-of-fit coefficient of the AR(1) model including the factor upon the nativeAR(1). Hence, figures above unity are favourable for the factor. Note that we display two sets of columns: the firstwith the factors built with the sectors most associated to the output gap and the second with the complimentaryfactors.
Table 2. Adjusted R-squared ratio between AR(1) factor-augmented and native AR(1) model (*)Factor: More sensitive sectors Factor: Less sensitive sectors
PCSI PFEI CCSI CFEI OFEI PCSI PFEI CCSI CFEI OFEIFactor: Commerce + IndustryPC 0.977 1.013 0.984 0.996 1.009 0.977 0.992 1.004 0.991 0.995DC 0.981 1.026 0.985 1.003 1.020 0.974 0.993 0.993 0.985 0.992NDC 0.950 1.030 0.956 0.990 1.019 0.956 0.989 0.980 0.979 0.992GFCF 0.981 1.026 1.008 1.022 1.034 0.998 1.007 1.014 1.031 1.037CWK 0.983 1.025 1.013 1.060 1.058 0.982 1.015 0.997 1.034 1.039MEQ 0.986 1.021 1.029 1.023 1.033 .018 0.999 1.078 1.054 1.060Factor: Commerce + Industry + TransportationPC 0.980 1.010 0.991 0.995 1.005 0.976 0.995 1.000 0.991 0.996DC 0.986 1.025 0.986 0.998 1.013 0.972 0.993 0.993 0.987 0.994NDC 0.951 1.024 0.959 0.984 1.009 0.955 0.992 0.980 0.983 0.998GFCF 0.985 1.019 1.010 1.023 1.033 0.990 1.011 1.015 1.032 1.038CWK 0.981 1.028 1.006 1.058 1.059 0.982 1.014 1.001 1.034 1.038MEQ 0.997 1.010 1.032 1.020 1.030 1.001 1.006 1.087 1.059 1.066Factor: Commerce + Industry + Construction + Personal ServicesPC 0.980 0.990 1.002 0.990 0.992 0.980 1.017 0.987 0.997 1.010DC 0.971 0.995 0.991 0.987 0.994 0.988 1.018 0.988 0.998 1.012NDC 0.959 0.978 0.975 0.981 0.989 0.958 1.052 0.956 0.986 1.021GFCF 0.994 1.007 1.016 1.033 1.038 0.981 1.023 1.003 1.022 1.033CWK 0.986 1.020 0.997 1.034 1.039 0.981 1.015 1.009 1.058 1.056MEQ 1.010 1.000 1.078 1.061 1.066 0.986 1.017 1.035 1.020 1.032Factor: Commerce + Construction + Personal ServicesPC 0.978 0.997 1.001 0.993 0.999 0.983 1.015 0.992 0.991 1.002DC 0.972 1.003 0.988 0.992 1.002 0.990 1.017 0.997 0.989 1.002NDC 0.953 0.991 0.969 0.989 1.004 0.965 1.046 0.972 0.971 0.998GFCF 0.990 1.014 1.017 1.035 1.042 0.985 1.015 0.995 1.013 1.020CWK 0.985 1.022 1.001 1.042 1.046 0.979 1.015 1.006 1.051 1.051MEQ 1.004 1.012 1.075 1.062 1.070 0.994 0.995 1.028 1.001 1.008
(*) Adjusted R-squared ratio=Adjusted R-squared with the factor/Adjusted R-squaredwithout the factor. Orange-shaded cells=Adjusted R-squared ratio greater than 1.
Green-shaded cells=Adjusted R-squared ratio greater than 1.05 (5% of adjustment gain).Sample: 2001.I-2014.IV (56 observations). Source: Authors’ calculations.
Three salient features are worth mentioning. First, whatever the sectors included in the factor either more corre-lated to the output gap or not, the PCSI classification virtually in none of the cases provides material improve-ments. Consequently, the reading of this survey when including personal current situation questions (Q3, Q4, andQ5a) does not come out with relevant information for neither prospective consumption nor investment. Second,a comparison between a given factor with its complementary version is always favourable for the sectors moreakin to the output gap. However, an exception is made with the CICS sectors (Commerce + Industry + Construction
9Recall that the adjusted goodness-of-fit coefficient corresponds to Adjusted-R2=1− [(1− R2)(T − 1)/(T − P− 1)], where R2 is the ratiobetween the sum of regression squared residuals and total squared residuals, T is the sample size, and P the number of estimated coefficientsin the regression. Thus, the measure already considers the adjustment controlling for the number of regressors. This is particularly relevant toavoid the issue of overfitting.
13
+ Personal Services) in which 13 (out of 30) cases compared to 18 cases improves the model’s adjustment. This isimportant because this is the factor containing more information than the others and, despite this advantage, isunable to track better, specifically, DC and NDC. Third, there are two factors providing a satisfactory performanceindependently of the sectors considered and working well with all variables: PFEI and OFEI. As will be shownlater, these two specifications come out as the best for forecasting.
Our second in-sample result is regarding the statistical significance of the null hypothesis NH: φ1=...=φ4=0 ofequation (3). In this case, note that we test for this specific hypothesis being a number of different ways to considerit in terms of factor specification (level or differences), number of terms included in the regression, and the nativemodel by itself. Recall that we set this exercise obeying to a predictive criterion which has led to the best resultswith 4-term factor in levels with an AR(1) native model. Naturally, testing a particular economic theory willrequire a richer environment, such as that of Acuña, Echeverría, and Pinto-Gutiérrez (2020) to control for otherpossible effects occurring simultaneously. Having reported that there are numerous ways to carry out this typeof test, the issue of short sample is also latent. As the estimation sample comprises 56 observations (2001.I-2014.IV)and the evaluation sample 19 observations (2015.I-2019.III), we consider the full sample (2001.I-2019.III). Although75 observations could still be moderate to estimate six highly-correlated regression coefficients, we opt to conductthis exercise to disentangle and discriminate better between the proposed factors.
The results are presented in Table 3. First, in this exercise there are overall, fewer cases in which the factor comesout statistically significant compared to the previous case in which it may enhance the adjustment of the model.In this sense, this exercise could be considered as a refinement of the previous one and, thus, helping to focusin the most promising cases. Second, within those promising cases there is now a marked difference betweenPFEI and OFEI specifications, plainly favouring the OFEI index with sectors more related to the business cycle,and particularly for investment series. Third, following this exercise, the information from CITS sectors turnsspecifically useful to explain the dynamic of the investment series only (particularly MEQ), which is added to theprevious results of adjustment improvements.
In sum, PFEI and OFEI are the most fruitful re-classifications to explain the dynamic of the series considered,whereas in terms of the economic sectors in which the consumers participate, there is not a single specificationthat comes out as systematically superior—which will be determined predictively. Finally, there is supportingevidence for sectors more related to the business cycle than the remaining ones when explaining considered macroaggregates.
4.2 Out-of-sample results
Our first out-of-sample result concerns our first hypothesis, i.e. the extent to which consumer sentiment factorsbetter forecast macro aggregates and their components compared to the native un-extended chosen AR(p) model.The results are presented in Table 1A in Annex A.
Three facts are worth mentioning. First, although showing some important predictive gains (=1-RMSFE ratio%)specially at longer horizons, both PC and CWK do not deliver statistically significant results. For PC the best factorspecification is the PCSI compounded by the CICS sectors with the largest gain surrounding 18% at h=4. In turn,for CWK, there is not a single specification actually showing a predictive gain at the nowcast horizon, whereas forthe remaining horizons the CFEI specification with CI economic sectors comes out as the best forecasting strategy,with gains ranging from 4.1% to 11.3% at h=2 and 4, respectively.
Second, for DC and MEQ, there are noticeable predictive gains for all horizons, but statistically significant only ath=1 for DC, and at h=4 for MEQ. For DC, two cases provide significant results, using CFEI and OFEI both usinginformation from CIT sectors. Note that predictive gains coming from including the Transportation sector as thecomparison with CI sectors show a non-statistically significant, poorer performance. Also, the maximum gainwith CFEI achieves a material figure close to 14% (being 9.5% with the second-best alternative, OFEI). For MEQ,there are four statistically significant cases, all of them at h=4. Remarkably, all these results are obtained with theCFEI specification with each of the four economic sectors groups. Furthermore, there is virtually a tie betweenCI and CIT sectors (with gains of 7.8% and 7.7%), then a better performance is noticed with CICS sectors (8.5%),
14
and the best result is obtained with CCS (10.3%). Thus, it is important to note that the only factor that excludesconsumers working in Industry comes out as the best and, at the same time, by making pairwise comparisons,Construction seems to contribute more to the factor’s forecast ability.
Table 3. F-test results (p-value) of the consumer-based factor statistical significance (*)Factor: More sensitive sectors Factor: Less sensitive sectors
PCSI PFEI CCSI CFEI OFEI PCSI PFEI CCSI CFEI OFEIFactor: Commerce + IndustryPC 0.965 0.084 0.766 0.239 0.071 0.935 0.477 0.203 0.500 0.313DC 0.635 0.038 0.674 0.226 0.070 0.931 0.571 0.558 0.556 0.413NDC 0.976 0.042 0.875 0.209 0.048 0.872 0.430 0.592 0.429 0.273GFCF 0.971 0.105 0.380 0.127 0.028 0.756 0.334 0.161 0.103 0.076CWK 0.961 0.163 0.362 0.014 0.020 0.951 0.190 0.630 0.103 0.034MEQ 0.836 0.125 0.197 0.204 0.059 0.318 0.475 0.020 0.097 0.060Factor: Commerce + Industry + TransportationPC 0.923 0.091 0.525 0.293 0.102 0.971 0.411 0.317 0.480 0.268DC 0.504 0.028 0.612 0.345 0.148 0.981 0.561 0.564 0.488 0.348NDC 0.969 0.126 0.845 0.340 0.111 0.851 0.345 0.600 0.332 0.191GFCF 0.916 0.177 0.333 0.144 0.049 0.859 0.268 0.146 0.094 0.055CWK 0.986 0.124 0.513 0.023 0.014 0.962 0.239 0.508 0.090 0.035MEQ 0.697 0.230 0.170 0.213 0.083 0.446 0.418 0.010 0.095 0.048Factor: Commerce + Industry + Construction + Personal ServicesPC 0.802 0.529 0.164 0.515 0.364 0.900 0.063 0.750 0.242 0.065DC 0.981 0.545 0.581 0.521 0.343 0.431 0.032 0.553 0.307 0.157NDC 0.766 0.584 0.602 0.397 0.351 0.918 0.030 0.927 0.262 0.034GFCF 0.761 0.311 0.160 0.076 0.048 0.971 0.146 0.406 0.170 0.070CWK 0.847 0.099 0.637 0.086 0.023 0.986 0.331 0.439 0.027 0.030MEQ 0.408 0.481 0.026 0.066 0.037 0.794 0.176 0.128 0.251 0.104Factor: Commerce + Construction + Personal ServicesPC 0.932 0.269 0.144 0.365 0.180 0.882 0.071 0.563 0.400 0.165DC 0.965 0.308 0.647 0.382 0.209 0.364 0.047 0.224 0.499 0.373NDC 0.904 0.347 0.660 0.282 0.176 0.858 0.081 0.562 0.511 0.162GFCF 0.886 0.244 0.150 0.070 0.034 0.947 0.270 0.624 0.263 0.147CWK 0.899 0.069 0.553 0.049 0.013 0.992 0.395 0.465 0.056 0.060MEQ 0.573 0.346 0.028 0.058 0.026 0.720 0.467 0.141 0.426 0.254
(*) p-value of the null hypothesis: NH: φ1=...=φ4=0 (see equation 1) against the alternativeof at least one term is equal to zero. Green-shaded cells=p<10%. Orange-shaded
cells=p<15%. Sample: 2001.I-2019.III (75 observations). Source: Authors’ calculations.
Finally, the most fruitful results are obtained for NDC and GFCF. A common result between them is that no singlefactor is significant at the nowcast horizon, and this also is extended to NDC at h=2. However, for NDC, notablythe predictive gains at h=3 occurred all with the same PCSI factor using the four different groups of economicsectors. Moreover, the best results are obtained with CICS (20.9%) and CCS (19.1%) sectors. Consequently, the keyfeature leading to these results is the inclusion of the answers of consumers working in Personal Services, which iscorroborated with the results at h=4, where aforementioned two factors—PCSI-CICS and PCSI-CCS—comes outas the best forecasting alternatives with predictive gains of 24% and 20.9%, respectively.
Regarding GFCF, despite some important gains with CI and CIT sectors, none of them comes out as statisticallysignificant. In contrast, using CICS and CCS sectors the results with OFEI are always significant for h>1. It isimportant to remark that only in this case the best results in terms of the lowest RMSFE ratio are not statisticallysignificant, but the second-best alternative actually is so at conventional levels of significance. These are the casesof CFEI and OFEI indexes generated with CICS information comparing 0.862 versus 0.941 at h=2, 0.829 versus0.958 at h=3, and 0.810 versus 0.951 at h=4, where OFEI results are always statistically significant. The same isobtained with CCS when comparing CFEI with OFEI at h=2: 0.855 versus 0.923, h=3: 0.809 versus 0.935, and h=4:
15
0.787 versus 0.928; all cases statistically significant with OFEI. Overall, however, for each considered case, the bestresults are always favourable to CCS sectors. In consequence, when considering the path of best results accordingto the lowest RMSFE ratio, this is composed by CFEI with CICS sectors at h=1, and CFEI with CCS sectors forh={2,3,4} despite the results of the CW test, but with a slightly different RMSFE ratio.
Regarding our second hypothesis, i.e. the extent to which consumers working in economic sectors more exposedto the business cycle provides better forecasting ability than those consumers working in the remaining sectors.The results are presented in Table 2A in Annex A.
The first salient feature is that a number of predictive gains are noticed favouring our hypothesis, but only a feware statistically significant. There is an overwhelming difference favouring the use of country-wide informationin the construction of the factor using either current sentiment or future expectation questions, plus the overallfuture expectation built jointly with the personal future expectation. On nine occasions, the OFEI indicator comesout as significant, and eight for CCSI, seven for CFEI, and two for PCSI. In terms of the economic sectors, theresults are spread across the four specifications, but the statistically significant cases are concentrated in CICSsectors (12), then CIT (8), and CI and CCS (3 cases each). Thus, our second hypothesis is mostly supported for DC,NDC, and GFCF using CFEI and OFEI factors with information coming from CICS sectors. These results, addedto the results of Table 1A, suggest a superior forecast ability of the CICS sectors over remaining ones when usingOFEI factor for GFCF, and reinforce the use of PCSI factor with the same sectors for NDC.
Given the number of specifications and results, in Table 4 we provide a summary of the best forecasting path ac-cording to the smaller RMSFE ratio for each variable. Note that in some cases, the smallest RMSFE ratio is achievedwith a non-statistically-significant setup, despite having available a second-best alternative with a slightly higherRMSFE ratio but rejecting the null hypothesis proposed by the CW test. This table highlights the importance ofthe PCSI specification for consumption, and the CFEI for investment series. It also makes clear that the most fruit-ful results in terms of predictive accuracy are found for NDC and GFCF, showing smaller RMSFE ratios. Finally,CICS and CCS sectors boldly provide the most valuable information for forecasting any of the considered macroaggregates.
4.3 Robustness results
In this subsection, we report the results of an extensive robustness exercise consisting in different twists to thebaseline setup. In particular, we assume different native best AR forecasting models, which is particularly relevantfor both DC and MEQ, as well as a different number of factor terms included in the regression. Notice that for DCand MEQ, according to Figure 4, the AR(4) model comes out as the best for some specific horizons (for all casesexcept MEQ at h=3 and 4). Unlike the baseline setup, we also consider the factor in levels. All these exercises areconducted to find the most fruitful way to consider the factor with an out-of-sample objective. Thus, the aim is tochallenge the results of Table 4.
All detailed results are presented in the Online Appendix, mimicking the results of Table 2 (Adjusted R-squaredratio), Table 3 (F-test results), Table 1A (RMSFE ratio between the native AR(1) and the factor-augmented ARmodel), and Table 1B (RMSFE ratio between the AR model augmented with the more sensitive factor and the ARmodel augmented with the less sensitive factor to the business cycle), but we provide a summary to ease compar-isons. Also, as we try the AR(2), AR(3), and AR(4) as native models, there is a new type of results when comparingto the native AR(1). However, these results are shown for reference only, as they are not fully comparable to theAR(1) because the factor-augmented AR(p) with p={2,3,4} could exhibit a better performance not exclusively tothe information contained in the factor but to a "larger p" (a model more suitable for the variable at hand).
Hence, the Online Appendix is organised according to the following scheme:
• A1: Native model: AR(2), factor specification: ∆x, number of factor terms: 4
• A2: Native model: AR(3), factor specification: ∆x, number of factor terms: 4
• A3: Native model: AR(4), factor specification: ∆x, number of factor terms: 4
16
Table 4: Summary of the best forecasting baseline setupby macro aggregate (*)
Setup h=1 h=2 h=3 h=4PC RMSFE: 0.983 0.958 0.874 0.818
Sectors: CICS CICS/CCS CCS CICSIndex: PCSI PCSI PCSI PCSI
DC RMSFE: 0.863? 0.937 0.886 0.856Sectors: CIT CIT CCS CCSIndex: CFEI CFEI CCSI CCSI
NDC RMSFE: 0.958 0.872 0.791? 0.760??
Sectors: CICS CICS CICS CICSIndex: PCSI PCSI PCSI PCSI
GFCF RMSFE: 0.945 0.855† 0.809† 0.787†
Sectors: CICS CCS CCS CCSIndex: CFEI CFEI CFEI CFEI
CWK RMSFE: 1.002 0.959 0.983 0.887Sectors: CCS CI CCS CIIndex: CFEI CFEI OFEI CFEI
MEQ RMSFE: 0.992 0.980 0.976 0.897?
Sectors: CICS CICS CCS CCSIndex: PFEI CFEI CFEI CFEI
(*) “RMSFE” stands for Relative RMSFE1h (see equation 3). “Sectors” stands
for the economic sectors considered in the consumer-perception-based factor(see subsection 3.1). “Index” stands for the timing (current/ future) and subject(personal/country-wide) feature of the factor. (†) For this case there is available
a statistically significant result but with a slightly higher RMSFE ratio. Clarkand West (2007) test’s p-value: (???) p<1%, (??) p<5%, and (?) p<10%.
Sample: 2015.I-2019.III (19 observations). Source: Authors’ calculations.
• A4: Native model: AR(1), factor specification: ∆x, number of factor terms: 1
• A5: Native model: AR(1), factor specification: ∆x, number of factor terms: 2
• A6: Native model: AR(1), factor specification: ∆x, number of factor terms: 3
• A7: Native model: AR(1), factor specification: x, number of factor terms: 1
• A8: Native model: AR(1), factor specification: x, number of factor terms: 2
• A9: Native model: AR(1), factor specification: x, number of factor terms: 3
• A10: Native model: AR(1), factor specification: x, number of factor terms: 4.
Consequently, tabulated results are shown according to:
• Baseline Tables 2 and 3 (Adjusted R-squared ratio and F-test results) = merged to Tables 1.A1 to 1.A10,
• Baseline Table 1A (RMSFE ratio between the native AR(1) and the factor-augmented AR model) = Tables2.A1 to 2.A10,
• New tables comparing AR(p), p={2,3,4}models with the AR(1) = Tables 3.A1 to 3.A3,
• Baseline Table 2A (RMSFE ratio between factor-augmented AR models according to its relationship with thebusiness cycle) = Tables 4.A1 to 4.A10.
17
Thus, for instance, for all series in which Figure 4 show that the best model is the AR(1), the comparisons should bemade across Table 2.A4 to Table 2.A10. In turn, for those series where the best model is the AR(4), the strength ofthe factor should be looked up in Table 3.A3. In this sense, and recalling that the primary interest is in forecastingand the comparison between sector, the primary conducting results are obtained from Tables 2.A4-2.A10 (or 3.A3)and Tables 4.A4-4.A10, with the in-sample results (Tables 1.A1 to 1.A10) analysed at the end.
The summary of robustness exercises is presented in Table 5. Notice that unlike Table 4, we consider two types ofresults, those statistically significant according to the CW test, and those not statistically significant. The latter areincluded to have an appraisal on the predictive gains on a small sample environment, where the CW test may notbe the most suitable but we use it as we have no better alternative, as well as to make it comparable to Table 4.
For PC, there are noticeable predictive gains for all horizons when comparing to baseline results, except at h=3in the non-statistically significant case. At h=1, there is a significant gain in neither the baseline nor alternativeestimations. However, the 2-term factor in levels with CIT sectors and OFEI index deliver better results as well asshowing better in-sample diagnostics. At h=2, there is a superior result in both RMSFE precision and significance.The results with the consumer-factor in levels including one and three terms, significant and non-statisticallysignificant, respectively, outperform the baseline results. More remarkably, the significant predictive gain of 4.9%with CIT sectors and CCSI index comes out as a better alternative than the remaining sectors less exposed to thebusiness cycle, and with better diagnostics. At h=3, the newly found statistically significant result does not providea greater predictive gain than the non-significant case, and its diagnostics are also non-supportive. For h=4, thebest results are obtained with the 4-term factor in levels, and its precision only improves for the non-significantcase. However, it is important to remark that robustness exercises find a superior statistically significant role forthe consumer-factor at any horizon (unlike the baseline results), except the nowcasting one, in which only onevariable exhibits a significant gain.
For DC, we present the results assuming that the best forecasting model is, first, the AR(1) and second the AR(4),which is in line with Figure 4. When assuming the AR(1) as the best model, the baseline results are still the bestperformers across all horizons, despite that for h>1 there are now statistically significant cases but with a smallerpredictive gain. The competing setup always includes a 2-term factor in levels. However, all these results for DCare due to the room opened by the use of the second-best forecasting model, the AR(1), and thus, it is relevantto delve in the results obtained with the AR(4). In this case, despite not finding statistically significant results, allpredictive gains are smaller than the AR(1) case, which is expected, but predictive gains nonetheless. Remarkably,all these results are obtained with the same sectors (CIT) and with two indexes: CFEI and CCSI. The best resultsare always obtained with the factor in first differences.
For NDC, robustness results provide enhanced predictive gains for h=1 and h=4. The case of NDC is the only onein the entire exercise showing a statistically significant predictive gain at the nowcasting horizon, which is also ofa material size (25.9%), and compounded by sectors (CCS) which are statistically superior to the remaining onesless exposed to the business cycle; and the PFEI index. For intermediate horizons, the results are mixed. In 8 out of9 cases the best setup is obtained with the factor in levels, and with a relatively high number of terms. For h=2, thestatistically significant result is obtained with CCS sectors and the PCSI index, which along with PFEI, are the twomost competitive indexes for this macro variable. Unlike the baseline case, for h=2 a statistically significant gainis found, completing all horizons with significant results. However, for these significant cases, the diagnostics arenot supportive, despite being superior at forecasting than the remaining sectors less exposed to the business cycle.Also, no clear pattern of sectors across horizons is found. At h=3 the baseline result remains the best, whereas h=4yields slightly better results (0.698 versus 0.760) with the factor in levels and replacing CICS by CI sectors.
For GFCF, none of the robustness results provides improvements. Thus, despite that the path of best results acrosshorizons does not provide statistically significant results, the significant ones show smaller gains also with the 4-term factor in first differences. From an in-sample point of view, all these results are very promising and favouringour hypotheses. Also, the out-of-sample comparison with sectors less exposed to the business cycle favours oursecond hypothesis, but being non-significant according to the HLN test.
For CWK, the situation with robustness exercises improves just a bit. At h=1, the baseline result with CICS sectorsand CFEI index show inexistent predictive gains, but using the factor in levels delivers a predictive gain of 4.9%
18
(statistically non-significant) with same sectors but PCSI index. A similar situation is found at h=2 and 3, wheremild, statistically non-significant gains are obtained with different sectors, indexes, and consumer-factor setups.At h=4, the baseline result persists as the best with CI sectors and CFEI index, and with in-sample diagnosticsfavouring our hypotheses.
Finally, for MEQ we analyse the results assuming that the best model according to Figure 4 is the AR(4), whichclearly applies for h={1,2}, but is not the case at h={3,4}—where the AR(1) comes out as the best. When assumingthat the best model is the AR(1), the results tend to favour our first hypothesis, in the sense that using the factorin levels improves the predictive gains for h={1,2,3}. However, they do not support our second hypothesis, andsectors less exposed to the business cycle could be as good as those more exposed to it. At h=4, in turn, thebaseline result remains as the best option, corresponding to a statistically significant predictive gain of 10.3%with CCS sectors and CFEI index that, and at the same time, is better than the remaining sector less exposed tothe business cycle. Also, in-sample diagnostics give support to our out-of-sample results. When assuming theAR(4) as the best model, we find statistical gains for all horizons, despite not statistically significant. The resultsfor h={1,2} are more relevant in the light of Figure 4, and in both cases the RMSFE with CICS sectors and PFEIindex gives support to our hypotheses–first, showing predictive gains of 5.2% and 6% for h={1,2} and, second,predicting better than do sectors less exposed to the business cycle.
In sum, from our baseline setup we conclude that PCSI along with CCSI indexes provide the most fruitful resultsfor consumption series, and CFEI doing so for investment series. In turn, NDC and GFCF are the variables showingthe greatest predictive gains of the whole exercise. Lastly, CICS and CCS sectors undoubtedly lead to the bestforecasting results as well as better forecasts compared to remaining sectors less exposed to the business cycle.From the robustness exercises, we confirm that NDC and GFCG are the most benefited series when using theconsumer-based factor for forecasting as well as the pre-eminence of CCS and CICS sectors. However, a majorshift in favour of the OFEI index is found, which in our baseline setup leads to the best result in just one case,whereas in the alternative setups comes out as the best alternative for investment series. This shift is due tothe different way in which the factor is included in the forecasting regression. In this sense, the factor in levelscomes out as the best alternative at shorter horizons, i.e. h={1,2}, whereas the first differences are superior for theremaining horizons.
5 Concluding remarks
The aim of this article is two-fold: (i) to analyse the extent to which consumer sentiment indexes better forecastaggregate consumption and its components, by constructing new indexes using labour market characteristics ofrespondents using a unique dataset, and re-focusing aggregate answers according to timing and subject, and(ii) to analyse to what extent a greater exposure of consumers/workers to the business cycle comes out with abetter ability to nowcast and short-term forecast consumption, compared to consumers/workers that participatein economic sectors less exposed to the business cycle. Our claim assumes that consumers work in sectors wheredecisions are taken depending to a greater or lesser degree on the current and future economic situation. Theconsumers can take the decisions themselves, be aware of them, or be affected by their consequences. Thus, weassume the existence of a learning mechanism permeable to the sentiment of workers that is present when they aresurveyed as consumers. We call this mechanism "exposure" and we operationalize it by means of the correlationof worker’s economic sector with the output gap. Since this argument is extensible to investment decisions, weinclude total investment and its components in the analysis.
Despite the considerable amount of related literature, this article goes one step further by using microdata evi-dence of consumer sentiment predictability over aggregate consumption and investment, taking advantage of aspecific survey on consumers’ perceptions which, at the same time, is linked to a survey of employment/ un-employment of the Greater Metropolitan area of Chile’s capital city. These are the Survey of Employment andUnemployment in Greater Santiago and the Survey of Perception and Expectations on the Economic Situation inGreater Santiago, both elaborated by the Universidad de Chile’s Centro de Microdatos.
19
Table 5: Summary of the best forecasting path by macro aggregate under different setups (*)h=1 h=2 h=3 h=4
Stat. s.1 Not stat. s.1 Stat. s.1 Not stat. s.1 Stat. s.1 Not stat. s.1 Stat. s.1 Not stat. s.1
Private Consumption (best native model2: AR(1))RMSFE3 - 0.837 0.951? 0.890 0.974? 0.874 0.914? 0.754Sectors4 - CIT CIT CICS CICS CCS CCS CIIndex5 - OFEI CCSI PFEI PCSI PCSI PCSI PCSIHLN test6 - 1.034 0.935 0.997 0.983 1.021 0.925 0.770Factor7 - x x x ∆x ∆x x xNo. of terms8 - 2 1 3 2 4 4 4Adj. R-sq. (more sens.)9 - 1.011 1.012 1.005 0.994 0.978 0.997 1.014Adj. R-sq. (less sens.)9 - 1.014 1.007 1.043 0.995 0.983 1.002 0.990F-Test (more sens.)10 - 0.122 0.075 0.618 0.813 0.932 0.645 0.215F-Test (less sens.)10 - 0.106 0.144 0.072 0.798 0.882 0.568 0.846
Durable Consumption (best native model2: AR(1))RMSFE3 0.863? 0.912 0.941?? 0.937 0.957? 0.886 0.955? 0.856Sectors4 CIT CI CI CIT CIT CCS CI CCSIndex5 CFEI CCSI OFEI CFEI OFEI CCSI OFEI CCSIHLN test6 0.922 0.961 0.976 0.952? 0.975 0.985 0.969 0.980Factor7 ∆x x x ∆x x ∆x x ∆xNo. of terms8 4 2 2 4 2 4 2 4Adj. R-sq. (more sens.)9 0.998 1.012 1.027 0.998 1.023 0.988 1.027 0.988Adj. R-sq. (less sens.)9 0.987 1.006 1.017 0.987 1.018 0.997 1.017 0.997F-Test (more sens.)10 0.345 0.162 0.022 0.345 0.031 0.647 0.022 0.647F-Test (less sens.)10 0.488 0.142 0.083 0.488 0.079 0.224 0.083 0.224
Durable Consumption (best native model2: AR(4))RMSFE3 - 0.885 - 0.978 - 0.928 - 0.919Sectors4 - CIT - CIT - CIT - CITIndex5 - CFEI - CFEI - CCSI - CCSIHLN test6 - 0.903? - 0.936? - 0.947 - 0.964Factor7 - ∆x - ∆x - ∆x - ∆xNo. of terms8 - 4 - 4 - 4 - 4Adj. R-sq. (more sens.)9 - 0.997 - 0.997 - 0.990 - 0.990Adj. R-sq. (less sens.)9 - 0.990 - 0.998 - 0.999 - 0.999F-Test (more sens.)10 - 0.411 - 0.411 - 0.476 - 0.476F-Test (less sens.)10 - 0.550 - 0.550 - 0.134 - 0.134
Non-durable Consumption (best native model2: AR(1))RMSFE3 0.741?? 0.769 0.896?? 0.653 0.791? 0.742 0.698? 0.700Sectors4 CCS CICS CCS CCS CICS CCS CI CITIndex5 PFEI PFEI PCSI PFEI PCSI PFEI PCSI PCSIHLN test6 0.831?? 0.809 0.870?? 0.932 0.866 0.964 0.697?? 0.694??
Factor7 x x x x ∆x x x xNo. of terms8 3 4 2 4 4 4 4 4Adj. R-sq. (more sens.)9 1.025 1.016 0.993 1.016 0.959 1.016 0.979 0.974Adj. R-sq. (less sens.)9 1.077 1.062 0.996 1.062 0.958 1.062 0.966 0.969F-Test (more sens.)10 0.352 0.450 0.555 0.450 0.766 0.450 0.710 0.784F-Test (less sens.)10 0.122 0.220 0.470 0.220 0.918 0.220 0.806 0.733
20
Table 5 (cont.): Summary of the best forecasting path by macro aggregate under different setups (*)h=1 h=2 h=3 h=4
Stat. s.1 Not stat. s.1 Stat. s.1 Not stat. s.1 Stat. s.1 Not stat. s.1 Stat. s.1 Not stat. s.1
Gross Fixed Capital Formation (best native model2: AR(1))RMSFE3 - 0.945 0.923?? 0.855 0.935? 0.809 0.928?? 0.787Sectors4 - CICS CCS CCS CCS CCS CCS CCSIndex5 - CFEI OFEI CFEI OFEI CFEI OFEI CFEIHLN test6 - 0.938 0.873 0.879 0.867 0.878 0.876 0.876Factor7 - ∆x ∆x ∆x ∆x ∆x ∆x ∆xNo. of terms8 - 4 4 4 4 4 4 4Adj. R-sq. (more sens.)9 - 1.033 1.042 1.035 1.042 1.035 1.042 1.035Adj. R-sq. (less sens.)9 - 1.022 1.020 1.013 1.020 1.013 1.020 1.013F-Test (more sens.)10 - 0.076 0.034 0.070 0.034 0.070 0.034 0.070F-Test (less sens.)10 - 0.170 0.147 0.263 0.147 0.263 0.147 0.263
Construction and Works (best native model2: AR(1))RMSFE3 - 0.951 - 0.950 - 0.944 - 0.887Sectors4 - CICS - CICS - CCS - CIIndex5 - PCSI - PCSI - CFEI - CFEIHLN test6 - 0.984 - 0.992 - 0.996 - 0.980Factor7 - x - x - ∆x - ∆xNo. of terms8 - 4 - 4 - 3 - 4Adj. R-sq. (more sens.)9 - 0.978 - 0.978 - 0.993 - 1.060Adj. R-sq. (less sens.)9 - 0.982 - 0.982 - 0.998 - 1.034F-Test (more sens.)10 - 0.848 - 0.848 - 0.433 - 0.014F-Test (less sens.)10 - 0.817 - 0.817 - 0.768 - 0.103
Machinery and Equipment (best native model2: AR(1))RMSFE3 - 0.976 0.937?? 0.933 0.928?? 0.922 0.897? 0.941Sectors4 - CICS CCS CICS CICS CICS CCS CITIndex5 - OFEI OFEI OFEI OFEI OFEI CFEI PFEIHLN test6 - 1.047 1.017 0.998 1.025 1.015 0.934 0.960Factor7 - x x x x x ∆x xNo. of terms8 - 2 2 3 2 3 4 2Adj. R-sq. (more sens.)9 - 1.002 1.003 1.092 1.002 1.092 1.062 0.998Adj. R-sq. (less sens.)9 - 0.998 0.996 1.056 0.998 1.056 1.001 1.002F-Test (more sens.)10 - 0.491 0.480 0.006 0.491 0.006 0.058 0.708F-Test (less sens.)10 - 0.639 0.691 0.025 0.639 0.025 0.426 0.641
Machinery and Equipment (best native model2: AR(4))RMSFE3 - 0.948 - 0.940 - 0.956 - 0.987Sectors4 - CICS - CICS - CICS - CITIndex5 - PFEI - PFEI - PFEI - PCSIHLN test6 - 0.869 - 0.916 - 0.959 - 0.993Factor7 - ∆x - ∆x - ∆x - ∆xNo. of terms8 - 4 - 4 - 4 - 4Adj. R-sq. (more sens.)9 - 0.987 - 0.987 - 0.987 - 0.997Adj. R-sq. (less sens.)9 - 0.980 - 0.980 - 0.980 - 1.012F-Test (more sens.)10 - 0.483 - 0.483 - 0.483 - 0.674F-Test (less sens.)10 - 0.652 - 0.652 - 0.652 - 0.159
(*) Notes to Table 5. (1) “Stat. s.” stands for statistically significant; “Not stat. s.” denotes non-statistically significant resultaccording to the Clark and West (2007) test. p-value: (???) p<1%, (??) p<5%, and (?) p<10%. (2) “Best native model” reports thebest forecasting AR(p) model according to Figure 4. For PC, NDC, GFCF, and CWK the best model is the AR(1) for all horizonswhereas for DC andMEQ it is the AR(4) for some specific horizons. (3) “RMSFE” stands for Relative RMSFEh
1 (see equation 3).(4) “Sectors” stands for the economic sectors considered in the consumer-based factor (see subsection 3.1): CI, CIT, CICS, and
21
CCS. (5) “Index” stands for the timing (current/future) and subject (personal/country-wide) feature of the factor: CCSI, CFEI,PCSI, PFEI, and OFEI. (6) “HLN test” stands for Relative RMSFEh
2 (see equation 3) and the results of the Harvey, Leybourne,and Newbold (1997) test between factor-augmented AR models according to its relationship with the business cycle. p-value:(???) p<1%, (??) p<5%, and (?) p<10%. Orange-shaded cells=Figures below 1. Green-shaded cells= Figures below 1 andstatistically significant. (7) “Factor” stands for the way in which the consumer-based factor enters the forecasting equation (seeequation 1) either in levels (x) or first differences (∆x). (8) “No. of terms” refers to the number of factor terms included in theforecasting equation (see equation 1). 1=x/∆x, 2=x x(-1)/∆(x) ∆(x(t-1)), 3=x x(-1) x(t-2)/∆(x) ∆(x(t-1)) ∆(x(t-2)), and 4=xx(t-1) x(t-2) x(t-3)/∆(x) ∆(x(t-1)) ∆(x(t-2)) ∆(x(t-3)). (9) “Adj. R-sq. (more sens.)” and “Adj. R-sq. (less sens.)” stands for theAdjusted R-squared ratio between factor-augmented and native AR models, for the factors built with sectors more and lesssensitive to the business cycle. Orange-shaded cells=Adjusted R-squared ratio greater than 1. Green-shaded cells=AdjustedR-squared ratio greater than 1.05 (5% of adjustment gain). (10) “F-Test (more sens.)” and “F-Test (less sens.)” stands for thep-value of the consumer-based factor statistical significance. NH: φ1=...=φp=0 (see equation 1) against the alternative of at leastone term is equal to zero. Green-shaded cells=p<10%. Orange-shaded cells=p<15%. Sample: 2015.I-2019.III (19 observations).Source: Authors’ calculations.
The key advantage is that it is possible to link, with no error, the perception of a consumer with her/his character-istics as a labour market participant. Thus, it is possible to exploit databases’ richness by re-grouping consumerperceptions indexes based on labour-market characteristics, aiming to improve their forecasting ability over macroaggregates and hypothesis testing. This treatment also aims to disentangle labour market characteristics more akinto the forecasting task.
Our first hypothesis comes out as a multi-horizon forecast comparison between a simple AR model for a givenmacro aggregate and its corresponding survey-based factor-augmentation version, using different computationsof consumer indexes. The second hypothesis consists in an out-of-sample comparison between two survey-basedfactors, one grouping indicators from consumers currently working in sectors more sensitive to the business cycle,and the other conformed by consumers working in the remaining sectors surveyed. We proxy the business cycleby means of an output gap estimation.
These hypotheses are tested for Private consumption, Non-durable consumption, and Durable consumption, plusGross fixed capital formation, Construction & works, and Machinery & equipment, using the quarterly samplecovering from 2000.I to 2019.III (75 observations). We also make use of a re-classification of answers based ontiming (current/future) and subject (personal/country). Remarkably, we are able to distinguish the economicsector in which the consumer is currently working, adding new dimensions to exploit on top of the previousre-classification.
Our in-sample results suggest that the adjusted goodness-of-fit coefficient of the AR model improves in most caseswhen including any of the consumer-based factors. In particular, the PFEI and the OFEI specifications improvefor almost all macro aggregates considering any sectorial definition of the factor. In contrast, the PCSI is lesshelpful in explaining the dynamics of any macro aggregate built with any sector. Most of these improvements arefound for GFCF and its components, andDC somewhat less. Consumers working in Commerce and Industry sectorsprovide more useful information than remaining sectors, as well as when adding consumers on Construction andPersonal Services sectors without manufacturing. When accounting for the statistical significance of the consumer-based factor, there is a concentration of significant cases using future-based indexes (PFEI, CFEI, and OFEI) for thesame macro aggregates that exhibit a better adjustment (GFCF and DC). Also, the evidence is most favourable forsectors mostly related to the business cycle–proxied by the output gap–rather than those that are less so or plainlyinsensitive to it.
Our out-of-sample results show, with a baseline added to alternative setups, predictive gains at all horizons forall variables, in some cases of a sizable magnitude and statistically non-significant in others. The most fruitfulresults are obtained for NDC and GFCF, which are statistically superior to the predictive benchmark model whenconformed by sectors more akin to the business cycle. There is not a silver-bullet definition of the factor helpingat the same time for all aggregates across all horizons. However, more often than not for consumption series, thePCSI followed by the CCSI lead to the best results. In turn, the OFEI do so for investment series. Regarding theeconomic sectors in which consumers participate and provide more useful predictive information, unquestionably
22
the best forecasting results are concentrated in Commerce, Industry, Construction, and Personal Services (thus, outof 3,060 individuals surveyed per quarter, this factor is composed by an average of 1,735 answers per quarter)and in Commerce, Construction, and Personal Services (comprising 1,313 answers per quarter on average). Whencomparing between factors, the results are largely favourable to the factors based on consumers more exposed tothe business cycle. In this line, the most fruitful results are obtained with Commerce, Industry, and Transportationsectors and Commerce, Construction, and Personal Services that outperform the remaining sectors less exposed tothe business cycle.
Although surveys of consumers perceptions provide rich information, these results are important as it is necessaryto disentangle ideally from the most disaggregate level to have a better revealing economic behaviour.
References
1. Abberger, K., 2007, "Forecasting Quarter-on-Quarter Changes of German GDP with Monthly Business TendencySurvey Results," Working Paper 40, ifo Leibniz Institute for Economic Research at the University of Munich.
2. Acuña, G., C. Echeverría, and C. Pinto-Gutiérrez, 2020, " Consumer Confidence and Consumption: Empiri-cal Evidence from Chile," International Review of Applied Economics 34(1): 75-93.
3. Albagli, E., J.A. Fornero, M.A. Fuentes, and R. Zúñiga, 2019, " On the Effects of Confidence and Uncertaintyon Aggregate Demand: Evidence from Chile," Economía Chilena 22(3): 8-33.
4. Basselier, R., D. de A. Liedo, G.L. Langenus, 2018, "Nowcasting Real Economic Activity in the Euro Area:Assessing the Impact of Qualitative Surveys," Journal of Business Cycle Research 14: 1-46.
5. Blanchard, O.J. and N.G. Mankiw, 1988, "Consumption: Beyond Certainty Equivalence," American EconomicReview 78(2): 173-177.
6. Box, G.E.P. and G.M. Jenkins, 1970, Time Series Analysis: Forecasting and Control, first edition, San Francisco,USA: Holden Day.
7. Bruno, G. and M. Malgarini, 2002, "An Indicator of Economic Sentiment for the Italian Economy," Documenti diLavoro 28, ISAE Istituto di Studi e Analisi Economica.
8. Bürgi, C. and T.M. Sinclair, 2015, "A Nonparametric Approach to Identifying a Subset of Forecasters that Outper-forms the Simple Average," Research Program on Forecasting, Working Paper 2015-006, The George Washing-ton University.
9. Calvo, G. and M. Ricaurte, 2012, "Indicadores Sintéticos para la Proyección de Imacec en Chile," Working Paper656, Central Bank of Chile.
10. Carroll, C.D., J.C. Fuhrer, and D.W. Wilcox, 1994, "Does Consumer Sentiment Forecast Household Spending?If So, Why?" American Economic Review 84(5): 1397-1408.
11. Centro de Microdatos, 2016, Diseño Muestral Encuesta de Ocupación y Desocupación en el Gran Santiago, Centrode Microdatos, University of Chile.
12. Chanut, N., M. Marcel, and C.A. Medel, 2019, "Can Economic Perceptions Surveys Improve MacroeconomicForecasting in Chile?," Economía Chilena 22(3): 34-97.
13. Clark, T.E. and M.W. McCracken, 2001, "Tests of Equal Forecast Accuracy and Encompassing for NestedModels," Journal of Econometrics 105: 85-110.
14. Clark, T.E. and M.W. McCracken, 2005, "Evaluating Direct Multistep Forecasts," Econometric Reviews 24: 369-404.
15. Clark, T.E. and K.D. West, 2007, "Approximately Normal Tests for Equal Predictive Accuracy in NestedModels," Journal of Econometrics 138(1): 291-311.
23
16. De Mello, E.P.G. and F.M.R. Figueiredo, 2014, "Assessing the Short-term Forecasting Power of Confidence Indices,"Working Paper 371, Banco Central do Brasil.
17. Diebold, F.X. and R.S. Mariano, 1995, "Comparing Predictive Accuracy," Journal of Business and EconomicStatistics 13: 153-263.
18. Doms, M. and N. Morin, 2004, "Consumer Sentiment, the Economy, and the News Media," Finance and Eco-nomics Discussion Series, Staff Working Paper 2004-51, Federal Reserve Board.
19. Figueroa, C. and M. Pedersen, 2019, "Extracting Information on Economic Activity from Business and Con-sumer Surveys in an Emerging Economy (Chile)," Economía Chilena 22(3): 98-131.
20. Fuhrer, J.C., 1993, "What Role Does Consumer Sentiment Play in the U.S. Economy?" New England EconomicReview January: 32-44.
21. Garnitz, J., R. Lehmann, and K. Wohlrabe, 2017, "Forecasting GDP all over the World: Evidence from Compre-hensive Survey Data," Working Paper 40, ifo Leibniz Institute for Economic Research at the University ofMunich.
22. Gayer, C., 2005, "Forecast Evaluation of European Commission Survey Indicators," Journal of Business CycleMeasurement and Analysis 2(2): 157-183.
23. Geiger, M. and J. Scharler, 2019, "How Do Consumers Assess the Macroeconomic Effects of Oil Price Fluctu-ations? Evidence from U.S. Survey Data," Journal of Macroeconomics 62: 103134.
24. Ghysels, E., D. Osborn, and P.M. Rodrigues, 2006, Forecasting Seasonal Time Series, in G. Elliot, C.W.J. Granger,and A. Timmermann (Eds.), Handbook of Economic Forecasting, volume 1, North Holland: Elsevier.
25. Girardi, A., R. Golinelli, and C. Pappalardo, 2017, "The Role of Indicator Selection in Nowcasting Euro-areaGDP in Pseudo-real Time," Empirical Economics 53: 79-99.
26. Hansson, J., P. Jansson, and M. Löf, 2003, "Business Survey Data: Do They Help in Forecasting the Macro Econ-omy?," Working Paper 84, National Institute of Economic Research, Stockholm, Sweden.
27. Harvey, D., S. Leybourne, and P. Newbold, 1997, "Testing the Equality of Prediction Mean Squared Errors,"International Journal of Forecasting 13: 281-291.
28. Howrey, E.P., 2001, "The Predictive Power of the Index of Consumer Sentiment," Brookings Papers on EconomicActivity 2001(1): 175-207.
29. Lahiri, L., G. Monokroussos, and Y. Zhao, 2015, "Forecasting Consumption: The Role of Consumer Confi-dence in Real Time with Many Predictors," Journal of Applied Econometrics 31(7): 1254-1275.
30. Li, C., 2011, "Consumer Expectation and Output Growth: The Case of China," Economics Letters 113: 298-300.
31. Ludvigson, S., 2004, "Consumer Confidence and Consumer Spending," Journal of Economic Perspectives 18:29-50.
32. Moon, H. and J. Lee, 2012, Forecast Evaluation of Economic Sentiment Indicator for the Korean Economy, in Bankfor International Settlements (Ed.), Proceedings of the Sixth IFC Conference on Statistical Issues and Activitiesin a Changing Environment 36: 180-190.
33. Österholm, P., 2014, "Survey Data and Short-term Forecasts of Swedish GDP Growth," Applied EconomicsLetters 21(2): 135-139.
34. Riquelme, V. and G. Riveros, 2018, "Un Indicador Contemporáneo de Actividad (ICA) para Chile," EconomíaChilena 21(1): 134-149.
35. Shröeder, M. and F.P. Hüfner, 2002, "Forecasting Economic Activity in Germany: How Useful are Sentiment Indica-tors?" Discussion Paper 02-56, Zentrum für Europäische Wirtschaftsforschung GmbH / Centre for EuropeanEconomic Research.
24
36. Souleles, N.S., 2004, "Expectations, Heterogeneous Forecast Errors, and Consumption: Micro Evidence fromthe Michigan Consumer Sentiment Survey," Journal of Money, Credit and Banking 36: 39-72.
37. Survey of Consumers, University of Michigan, 2020, Survey Description, accessed on July 7, 2020.
25
AB
asel
ine
resu
lts:
AR
(1)m
odel
Tabl
e1A
.Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(1
)mod
el(*
)PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
006
1.17
5???
1.07
2?1.
015
1.08
8??
0.99
41.
178???
1.08
4??
1.01
51.
086??
0.98
31.
171?
1.12
6??
1.03
71.
095?
0.98
91.
164?
1.11
4??
1.02
91.
088?
20.
974
1.37
3??
1.08
61.
079
1.21
7???
0.96
11.
367??
1.10
4?1.
106
1.21
7???
0.95
81.
314??
1.15
2??
1.17
6?1.
223??
0.95
81.
314??
1.12
5??
1.15
8?1.
215??
30.
919
1.48
7??
1.01
81.
099
1.24
9??
0.90
41.
489??
1.01
61.
128
1.25
9??
0.88
01.
467??
1.05
61.
217?
1.28
9?0.
874
1.45
9??
1.03
51.
192?
1.27
5?
40.
947
1.46
9??
1.00
21.
119
1.22
2??
0.91
81.
472??
0.99
71.
143
1.23
5?0.
818
1.47
7??
1.04
41.
233?
1.28
2?0.
830
1.46
5??
1.02
91.
203?
1.26
7?
hD
urab
leC
onsu
mpt
ion
11.
004
0.98
30.
946
0.90
20.
932
0.99
20.
985
0.94
10.
863?
0.90
5?0.
996
0.99
70.
932
0.89
40.
934
0.99
30.
995
0.91
70.
889
0.93
42
1.00
50.
987
0.97
20.
941
0.95
41.
000
0.98
50.
961
0.93
70.
940
1.01
60.
977
0.96
40.
959
0.94
01.
011
0.98
00.
953
0.96
40.
949
31.
006
0.96
80.
898
0.97
00.
938
1.00
00.
965
0.88
70.
962
0.92
51.
021
0.96
20.
893
0.96
70.
924
1.01
60.
967
0.88
60.
969
0.93
24
1.00
90.
948
0.87
80.
945
0.89
31.
003
0.94
80.
867
0.93
40.
882
1.02
10.
950
0.86
00.
928
0.88
21.
018
0.95
30.
856
0.93
10.
891
hN
on-d
urab
leC
onsu
mpt
ion
11.
015
1.17
6???
1.06
10.
972
1.07
4??
1.00
61.
174???
1.05
70.
982
1.07
4??
0.95
81.
163??
1.06
01.
019
1.07
20.
961
1.15
3?1.
068
1.00
81.
067
20.
975
1.30
4??
1.08
90.
901
1.08
5??
0.95
61.
303??
1.06
60.
937
1.09
0?0.
872
1.29
5???
1.05
21.
038
1.11
8?0.
885
1.28
7??
1.04
51.
013
1.10
9?
30.
934???
1.36
0??
1.09
40.
859
1.07
80.
914???
1.36
3??
1.06
00.
886
1.08
40.
791?
1.38
1???
1.00
71.
016
1.12
70.
809??
1.36
7???
1.00
70.
989
1.11
84
0.94
71.
356???
1.05
70.
848
1.06
00.
925
1.36
2???
1.02
10.
875
1.06
80.
760??
1.41
4???
0.96
71.
009
1.12
90.
791??
1.39
6???
0.97
70.
984
1.12
3h
Gro
ssFi
xed
Cap
ital
Form
atio
n1
1.01
91.
001
1.06
70.
984
0.97
81.
025
1.02
21.
058
0.99
70.
997
1.05
11.
014
1.02
10.
945
v0.9
701.
049
1.00
01.
015
0.95
00.
965
21.
001
1.01
01.
015
0.91
60.
969
1.00
81.
036
1.01
80.
923
0.99
21.
047
1.01
90.
995
0.86
20.
941??
1.04
00.
998
0.98
70.
855
0.92
3??
31.
005
1.06
40.
983
0.85
00.
982
1.00
61.
083
0.97
90.
862
1.00
71.
034
1.06
00.
998
0.82
90.
958?
1.03
01.
042
0.99
20.
809
0.93
5?
40.
994
1.08
0?1.
009
0.83
30.
977
0.99
31.
090?
1.00
50.
842
1.00
21.
008
1.05
51.
036
0.81
00.
951?
1.00
51.
041
1.02
20.
787
0.92
8??
hC
onst
ruct
ion
and
Wor
ks1
1.03
4?1.
040
1.07
31.
002
1.03
11.
032??
1.04
71.
084
1.00
61.
032
1.02
21.
038?
1.07
91.
009
1.01
31.
023
1.03
6?1.
075
1.00
21.
009
21.
004
1.03
51.
043
0.95
91.
014
1.00
61.
040
1.05
50.
967
1.01
71.
009
1.03
21.
073
0.96
80.
997
1.00
71.
030
1.06
20.
960
0.99
13
0.99
51.
048
1.03
70.
910
1.01
50.
998
1.05
31.
039
0.91
91.
019
0.99
61.
041
1.08
30.
926
0.99
10.
995
1.04
01.
066
0.91
30.
983
40.
989
1.06
81.
066
0.88
71.
020
0.99
01.
071
1.06
60.
899
1.02
70.
979
1.05
3?1.
128??
0.91
00.
993
0.97
91.
053
1.10
3??
0.89
30.
984
hM
achi
nery
and
Equi
pmen
t1
1.01
31.
008
1.07
4?1.
026
1.00
21.
016
1.01
51.
068?
1.03
11.
013
1.03
50.
992
1.02
8?1.
002
1.00
41.
034?
0.99
41.
024
1.01
01.
004
20.
998
0.99
71.
023
1.02
01.
007
1.00
21.
017
1.02
81.
012
1.02
41.
036
1.01
40.
989
0.98
01.
009
1.03
01.
001
0.98
50.
986
0.99
93
1.01
51.
148?
1.03
30.
982
1.06
8?1.
013
1.15
4??
1.03
40.
989
1.10
0?1.
030
1.13
71.
035
0.99
11.
093?
1.02
91.
127
1.02
90.
976
1.06
14
0.99
31.
065
0.99
70.
922??
1.00
60.
995
1.05
90.
995
0.92
3??
1.02
51.
026
1.03
51.
012
0.91
5?1.
013
1.02
21.
037
1.00
80.
897?
0.99
5(*
)Rel
ativ
eR
MSF
E1 h(s
eeeq
uati
on3)
.Ora
nge-
shad
edce
lls=F
igur
esbe
low
1.G
reen
-sha
ded
cells
=Fig
ures
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Cla
rkan
dW
est(
2007
)tes
t’sp-
valu
e:(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
26
Tabl
e2A
:Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
fact
or-a
ugm
ente
dA
R(1
)mod
els
acco
rdin
gto
its
rela
tion
ship
wit
hth
ebu
sine
sscy
cle
(*)
PCSIPFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
001
1.07
20.
971
1.06
11.
079
0.98
21.
089
0.97
61.
048
1.07
70.
986
1.04
41.
046
1.05
91.
052
0.98
00.
986
1.02
91.
009
0.99
52
0.94
71.
114??
0.89
90.
895
1.00
00.
933
1.11
7??
0.91
60.
924
1.00
40.
879
1.02
20.
965
0.98
90.
981
0.88
31.
001
0.92
60.
962
0.95
93
0.95
91.
119?
0.87
80.
908
0.98
50.
949
1.12
4?0.
874
0.93
80.
997
0.83
11.
083
0.91
01.
046
1.00
80.
830
1.04
80.
882
1.01
90.
984
41.
012
1.06
40.
857
0.88
70.
932
0.98
41.
070
0.85
10.
911
0.94
50.
757?
1.06
80.
909
1.04
30.
992
0.77
31.
038
0.90
41.
009
0.97
7h
Dur
able
Con
sum
ptio
n1
1.02
50.
989
1.01
50.
997
0.99
31.
002
0.99
01.
008
0.91
60.
942
0.99
20.
994
0.97
90.
962
0.98
30.
989
0.97
50.
944
0.93
30.
965
21.
003
1.02
91.
042
0.96
11.
017
0.99
31.
028
1.02
70.
942
0.99
01.
025
1.00
31.
018
0.98
60.
983
1.02
11.
006
0.98
41.
011
1.01
23
1.01
61.
040
1.08
31.
015
1.03
71.
005
1.03
61.
064
0.99
31.
010
1.05
71.
026
1.05
70.
995
1.01
01.
047
1.03
01.
040
1.00
71.
041
41.
015
1.03
91.
090
1.04
11.
045
1.00
01.
040
1.06
91.
016
1.02
31.
048
1.03
61.
032
1.00
11.
032
1.03
61.
044
1.01
81.
004
1.06
9h
Non
-dur
able
Con
sum
ptio
n1
0.97
61.
097??
0.96
20.
943
1.02
60.
964
1.09
9??
0.95
80.
943
1.02
10.
928
1.06
00.
969
0.97
70.
978
0.91
21.
005
1.01
00.
953
0.94
32
0.96
11.
119???
0.91
80.
803
0.94
00.
944?
1.12
0???
0.89
30.
840
0.94
90.
807
1.10
2??
0.85
50.
959
0.96
80.
818
1.07
10.
856
0.93
10.
952
30.
949
1.12
7???
0.91
7??
0.76
4?0.
917
0.93
5?1.
132??
0.88
2??
0.79
1?0.
925
0.74
21.
156??
0.81
0?0.
966
0.96
30.
760
1.11
0?0.
816??
0.94
40.
953
40.
988
1.08
80.
907?
0.75
4?0.
885
0.97
31.
093?
0.86
8?0.
781?
0.89
30.
719?
1.17
7???
0.80
30.
987
0.96
70.
750?
1.13
5???
0.82
90.
976
0.97
1h
Gro
ssFi
xed
Cap
ital
Form
atio
n1
1.06
00.
998
0.99
51.
014
0.97
01.
063
1.02
60.
999
1.04
01.
004
1.12
61.
002
0.92
70.
934
0.94
61.
112
0.93
30.
896
0.92
10.
906
21.
007
0.93
50.
925
1.00
60.
922
1.01
00.
973
0.94
21.
041
0.97
21.
090
0.95
70.
922
0.88
00.
858
1.06
90.
896
0.90
10.
812
0.78
23
1.00
30.
896
0.84
9?0.
944
0.87
11.
004
0.92
40.
844?
0.99
20.
925
1.04
10.
889
0.90
5?0.
870
0.81
11.
039
0.83
30.
898
0.76
40.
728?
41.
031
0.89
10.
826??
0.90
50.
850?
1.03
30.
909
0.82
0??
0.95
40.
901
1.01
60.
852
0.89
7??
0.84
40.
793?
1.02
00.
809
.876???
0.74
30.
715??
hC
onst
ruct
ion
and
Wor
ks1
1.00
50.
984
0.95
61.
004
0.99
41.
004
0.99
70.
971
1.01
11.
000
0.98
20.
973
0.96
31.
016
0.95
20.
972
0.94
30.
955
0.98
40.
921
20.
989
0.98
50.
926??
0.99
60.
989
0.99
20.
994
0.94
0??
1.01
40.
998
1.00
00.
974
0.98
31.
016
0.94
60.
995
0.96
10.
965
0.97
50.
909
30.
993
0.98
90.
906??
0.98
10.
991
0.99
80.
996
0.90
3??
1.00
11.
004
0.99
90.
969
0.98
51.
015
0.93
60.
998
0.96
00.
954
0.96
00.
897
40.
994
0.97
4??
0.87
2??
0.94
80.
971
0.99
60.
979
0.86
5???
0.97
80.
991
0.96
70.
937
0.97
20.
999
0.91
1??
0.96
70.
929
0.92
4?0.
935?
0.87
2??
hM
achi
nery
and
Equi
pmen
t1
1.06
31.
098
1.06
11.
017
1.02
61.
060
1.11
11.
066
1.02
21.
039
1.13
11.
045
0.95
50.
978
1.02
31.
129
1.01
30.
930
0.99
31.
023
21.
020
0.99
90.
996
0.99
40.
955
1.01
81.
028
1.02
41.
001
0.98
21.
113
1.00
50.
967
0.93
50.
958
1.09
70.
964
0.94
60.
937
0.93
83
1.01
80.
975
0.86
00.
869
0.88
71.
004
0.98
00.
874
0.89
70.
923
1.03
20.
912
0.91
00.
908
0.90
51.
044
0.86
5?0.
892
0.88
50.
860
41.
044
0.98
40.
899
0.91
80.
923
1.04
10.
984
0.89
90.
937
0.95
11.
121
0.93
60.
941
0.92
20.
920
1.12
60.
909
0.93
00.
878
0.87
5(?
)Rel
ativ
eR
MSF
E2 h(s
eeeq
uati
on3)
.Ora
nge-
shad
edce
lls=F
igur
esbe
low
1.G
reen
-sha
ded
cells
=Fi
gure
sbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.H
arve
y,Le
ybou
rne,
and
New
bold
(199
7)te
st’s
p-va
lue:
(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
27
Does the Exposure to the Business Cycle Improve Consumer Perceptionsfor Forecasting? Microdata Evidence from Chile∗
Online Appendix
Fernando Faure†
Universidad de ChileCarlos A. Medel‡
Central Bank of Chile
September 1, 2020
Abstract
This is the Online Appendix of the Faure, F. and C.A. Medel (2020), "Does the Exposure to the Business CycleImprove Consumer Perception for Forecasting? Microdata Evidence from Chile" paper containing robustness exercises’results referred in subsection 4.3. These exercises are different variations of the baseline setup displayed accordingto the following scheme:
Factor In-sample Out-of-sample
No. Model Spec. Terms Adj. R-sq. F-test ARX/AR(1) ARX(p)/AR(p) ARX1/ARX2
Baseline AR(1) ∆y 4 Table 2 Table 3 Table 1A - Table 2A
1 AR(2) ∆y 4 Table 1.A1 Table 2.A1 Table 3.A1 Table 4.A1
2 AR(3) ∆y 4 Table 1.A2 Table 2.A2 Table 3.A2 Table 4.A2
3 AR(4) ∆y 4 Table 1.A3 Table 2.A3 Table 3.A3 Table 4.A3
4 AR(1) ∆y 1 Table 1.A4 Table 2.A4 - Table 4.A4
5 AR(1) ∆y 2 Table 1.A5 Table 2.A5 - Table 4.A5
6 AR(1) ∆y 3 Table 1.A6 Table 2.A6 - Table 4.A6
7 AR(1) y 1 Table 1.A7 Table 2.A7 - Table 4.A7
8 AR(1) y 2 Table 1.A8 Table 2.A8 - Table 4.A8
9 AR(1) y 3 Table 1.A9 Table 2.A9 - Table 4.A9
10 AR(1) y 4 Table 1.A10 Table 2.A10 - Table 4.A10
∗The views and ideas expressed in this paper do not necessarily represent those of the Central Bank of Chile or its authorities. Any error oromission remains at authors’ sole responsibility.
†BA in Economics and Administration undergraduate, Universidad de Chile. E-mail: [email protected].‡Economic Adviser to the Governor, Central Bank of Chile. E-mail: [email protected].
1R
obus
tnes
sre
sult
sI:
AR
(2)n
ativ
em
odel
,4-t
erm
fact
orin
diff
eren
ces
Tabl
e1.
A1.
Adj
uste
dR
-squ
ared
rati
obe
twee
nA
R(2
)fac
tor-
augm
ente
dan
dna
tive
AR
(2)m
odel
(*)
and
F-te
stre
sult
s(p
-val
ue)o
fthe
cons
umer
-bas
edfa
ctor
stat
isti
cals
igni
fican
ce(*
*)Fa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sFa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEI
Fact
or:C
omm
erce
+In
dust
ryPC
0.98
11.
024
0.99
01.
000
1.01
50.
984
0.99
91.
021
0.99
81.
002
0.96
80.
100
0.61
50.
277
0.05
30.
916
0.43
40.
017
0.44
60.
274
DC
0.99
01.
049
0.98
80.
996
1.02
10.
983
1.00
21.
004
0.98
60.
993
0.52
10.
001
0.62
90.
581
0.05
20.
954
0.25
10.
086
0.86
60.
575
ND
C0.
948
1.02
90.
955
0.98
91.
018
0.95
40.
986
0.97
70.
977
0.99
00.
982
0.04
60.
894
0.19
90.
049
0.88
10.
444
0.63
40.
432
0.28
4G
FCF
0.99
11.
005
1.02
01.
009
1.01
41.
009
0.99
91.
022
1.01
61.
019
0.76
50.
250
0.01
10.
277
0.09
10.
305
0.34
40.
063
0.22
00.
112
CW
K0.
995
0.99
91.
013
1.02
71.
019
1.00
21.
004
1.00
61.
017
1.01
60.
977
0.88
00.
365
0.06
00.
166
0.62
60.
672
0.50
20.
143
0.12
2M
EQ0.
993
1.00
81.
022
1.01
31.
022
1.01
90.
990
1.07
51.
035
1.03
80.
728
0.31
10.
096
0.43
60.
200
0.20
30.
555
0.01
70.
331
0.20
0Fa
ctor
:Com
mer
ce+
Indu
stry
+Tr
ansp
orta
tion
PC0.
983
1.01
71.
000
1.00
01.
011
0.98
31.
004
1.01
50.
998
1.00
20.
951
0.13
10.
237
0.29
00.
085
0.95
60.
302
0.04
70.
434
0.23
0D
C0.
994
1.04
70.
992
0.99
21.
012
0.98
31.
003
1.00
10.
986
0.99
40.
354
0.00
00.
345
0.73
70.
192
0.94
10.
280
0.15
00.
811
0.45
1N
DC
0.95
01.
022
0.95
80.
983
1.00
70.
953
0.99
00.
977
0.98
10.
996
0.96
40.
129
0.86
50.
327
0.11
00.
867
0.35
60.
637
0.33
90.
198
GFC
F0.
996
1.00
21.
025
1.00
91.
014
0.99
70.
998
1.02
31.
016
1.01
90.
702
0.32
00.
008
0.30
80.
129
0.53
80.
369
0.06
60.
181
0.08
4C
WK
0.99
41.
001
1.01
11.
027
1.02
11.
000
1.00
31.
008
1.01
71.
016
0.98
30.
854
0.44
10.
044
0.07
70.
680
0.71
80.
432
0.18
30.
190
MEQ
1.00
51.
000
1.02
81.
007
1.01
60.
995
0.99
21.
085
1.04
31.
047
0.65
60.
408
0.06
70.
515
0.29
00.
353
0.56
40.
010
0.28
00.
144
Fact
or:C
omm
erce
+In
dust
ry+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
990
0.99
41.
020
0.99
60.
996
0.98
41.
037
0.99
31.
002
1.02
10.
684
0.66
30.
013
0.50
50.
450
0.89
60.
014
0.62
80.
202
0.01
3D
C0.
981
1.00
11.
003
0.98
60.
993
0.99
01.
049
0.99
00.
993
1.01
50.
972
0.35
40.
086
0.87
30.
598
0.51
50.
000
0.56
70.
587
0.07
4N
DC
0.95
60.
976
0.97
30.
979
0.98
70.
957
1.05
10.
955
0.98
51.
019
0.78
30.
594
0.63
40.
399
0.35
60.
933
0.03
00.
942
0.25
90.
035
GFC
F1.
011
0.99
81.
024
1.01
81.
020
0.98
81.
004
1.01
31.
008
1.01
40.
265
0.39
40.
050
0.17
10.
086
0.87
60.
260
0.04
70.
341
0.13
5C
WK
1.00
71.
005
1.00
51.
018
1.01
50.
996
0.99
81.
012
1.02
51.
019
0.34
50.
587
0.55
60.
126
0.14
00.
952
0.93
70.
327
0.06
90.
129
MEQ
1.01
70.
991
1.07
51.
043
1.04
50.
987
1.00
21.
027
1.00
81.
018
0.21
00.
585
0.01
90.
262
0.14
60.
784
0.36
10.
092
0.54
40.
293
Fact
or:C
omm
erce
+C
onst
ruct
ion
+Pe
rson
alSe
rvic
esPC
0.98
71.
003
1.02
20.
999
1.00
40.
986
1.03
30.
994
0.99
71.
010
0.86
10.
307
0.00
90.
370
0.20
60.
901
0.01
20.
642
0.32
30.
090
DC
0.98
31.
012
1.00
10.
991
1.00
20.
991
1.05
20.
994
0.98
51.
002
0.93
10.
078
0.13
20.
702
0.31
20.
509
0.00
00.
389
0.84
90.
372
ND
C0.
951
0.98
90.
966
0.98
71.
002
0.96
51.
044
0.97
20.
970
0.99
60.
913
0.36
10.
692
0.28
10.
176
0.85
70.
082
0.64
30.
509
0.16
9G
FCF
1.00
51.
003
1.02
81.
017
1.02
20.
991
0.99
31.
001
1.00
31.
003
0.43
10.
347
0.03
70.
180
0.07
30.
840
0.66
40.
281
0.42
90.
326
CW
K1.
002
1.00
51.
007
1.02
11.
017
0.99
70.
997
1.01
11.
024
1.01
90.
603
0.64
50.
475
0.10
70.
134
0.94
70.
922
0.38
10.
059
0.09
6M
EQ1.
012
1.00
51.
074
1.04
41.
051
0.99
60.
977
1.01
60.
990
0.99
20.
322
0.44
20.
015
0.23
20.
108
0.67
60.
777
0.17
50.
743
0.61
0(*
)Adj
uste
dR
-squ
ared
rati
o=A
djus
ted
R-s
quar
edw
ith
the
fact
or/A
djus
ted
R-s
quar
edw
itho
utth
efa
ctor
.Ora
nge-
shad
edce
lls=A
djus
ted
R-s
quar
edra
tio
grea
ter
than
1.G
reen
-sha
ded
cells
=Adj
uste
dR
-squ
ared
rati
ogr
eate
rth
an1.
05(5
%of
adju
stm
entg
ain)
.Sam
ple:
2001
.I-20
14.IV
(56
obse
rvat
ions
).(*
*)p-
valu
eof
the
null
hypo
thes
is:N
H:φ
1=...
=φ
4=0
(see
equa
tion
1)ag
ains
tthe
alte
rnat
ive
ofat
leas
tone
term
iseq
ual
toze
ro.G
reen
-sha
ded
cells
=p<
10%
.Ora
nge-
shad
edce
lls=p<
15%
.Sam
ple:
2001
.I-20
19.II
I(75
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
1
Tabl
e2.
A1.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(2
)mod
el(*
)PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
254
1.32
8??
1.25
4?1.
236?
1.28
6??
1.23
91.
334??
1.26
31.
233
1.28
3?1.
183
1.35
3??
1.29
51.
244
1.29
4?1.
201
1.34
2?1.
297?
1.23
91.
288?
21.
351???
1.61
7???
1.39
0??
1.41
6??
1.50
7???
1.34
0??
1.61
4???
1.38
4??
1.42
3??
1.50
3??
1.28
3??
1.61
1???
1.38
4?1.
484??
1.52
3??
1.29
9??
1.60
2???
1.38
1?1.
471??
1.51
6??
31.
416???
1.79
3???
1.52
3??
1.52
0??
1.63
1???
1.40
4???
1.79
1???
1.49
8??
1.52
9??
1.62
9??
1.32
7??
1.80
0???
1.45
3?1.
622??
1.66
4??
1.34
0???
1.78
8???
1.46
4?1.
607??
1.66
0??
41.
489???
1.80
6???
1.58
4???
1.56
0??
1.65
2??
1.47
1???
1.80
7???
1.56
0??
1.56
9??
1.65
3??
1.36
3???
1.82
9???
1.51
9??
1.65
7??
1.69
4??
1.38
7???
1.81
2???
1.53
4??
1.64
4??
1.69
2??
hD
urab
leC
onsu
mpt
ion
11.
032
1.00
60.
975
0.95
50.
979
1.01
81.
008
0.97
50.
914
0.95
21.
016
1.03
00.
977
0.94
10.
981
1.01
41.
028
0.96
30.
934
0.98
0
20.
965
0.96
50.
945
0.91
30.
940
0.96
10.
970
0.93
60.
908
0.92
80.
970
0.96
80.
951
0.93
20.
930
0.96
40.
970
0.93
70.
934
0.93
5
30.
945
0.94
10.
866
0.92
10.
912
0.93
90.
940
0.85
80.
915
0.90
10.
948
0.93
70.
875
0.92
50.
901
0.94
30.
943
0.86
50.
923
0.90
7
40.
928
0.90
70.
839
0.88
20.
857
0.92
30.
911
0.83
10.
873
0.85
00.
931
0.91
20.
835
0.87
20.
849
0.92
70.
915
0.82
70.
869
0.85
5
hN
on-d
urab
leC
onsu
mpt
ion
11.
249?
1.33
6???
1.26
6???
1.22
2??
1.28
8???
1.23
8?1.
331???
1.26
6??
1.22
5?1.
285??
1.18
3?1.
344??
1.24
4?1.
247??
1.28
7??
1.19
5?1.
334??
1.26
1?1.
243?
1.28
4??
21.
344???
1.55
4???
1.47
0???
1.31
7??
1.43
5???
1.33
2???
1.55
1???
1.43
8???
1.33
9??
1.43
7???
1.25
3??
1.58
1???
1.36
4??
1.42
3??
1.46
8???
1.26
9??
1.57
0???
1.38
4???
1.40
7??
1.46
5???
31.
404???
1.65
9???
1.60
7???
1.37
6??
1.51
5???
1.38
9???
1.65
8???
1.56
7???
1.40
0???
1.51
8???
1.29
0???
1.69
8???
1.47
1???
1.51
3???
1.55
5???
1.30
9???
1.68
4???
1.49
5???
1.49
6???
1.55
5???
41.
477???
1.69
4???
1.65
0???
1.40
8??
1.55
0???
1.45
8???
1.69
6???
1.61
2???
1.43
6??
1.55
6???
1.33
8???
1.75
1???
1.53
0???
1.55
0???
1.59
9???
1.36
6???
1.73
2???
1.55
7???
1.53
3???
1.60
1???
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
065
1.03
9?1.
108??
1.03
21.
023
1.07
4?1.
060??
1.09
8??
1.04
41.
043?
1.10
7?1.
057?
1.05
8?0.
992
1.01
71.
103?
1.04
3?1.
053?
0.99
81.
012
21.
042
1.04
1?1.
060
0.96
51.
014
1.05
01.
062??
1.05
90.
971
1.03
71.
095
1.05
8??
1.02
70.
915
0.98
91.
087
1.03
9?1.
021
0.91
00.
972
30.
996
1.06
0?0.
996
0.82
30.
972
0.99
71.
078??
0.98
60.
837
1.00
41.
017
1.05
30.
997
0.80
80.
947
1.01
41.
036
0.99
10.
786
0.92
0?
41.
022
1.09
7??
1.03
90.
862
1.00
21.
024
1.10
7??
1.02
90.
869
1.03
01.
034
1.08
0??
1.05
0?0.
846
0.97
81.
032
1.06
6??
1.03
90.
828
0.95
5
hC
onst
ruct
ion
and
Wor
ks
10.
883
0.90
10.
915
0.87
90.
880
0.87
80.
903
0.91
80.
885
0.88
30.
873
0.87
70.
906
0.88
10.
868
0.87
30.
875
0.90
40.
880
0.86
5
20.
972
1.01
60.
991
0.96
70.
989
0.97
11.
018
1.00
10.
976
0.99
20.
972
0.99
60.
999
0.97
50.
979
0.97
10.
993
0.99
60.
972
0.97
5
31.
021
1.08
81.
017
1.01
01.
056
1.02
41.
091
1.02
51.
018
1.06
11.
022
1.06
91.
036
1.02
1?1.
045
1.02
21.
066
1.03
21.
015?
1.04
0
41.
054
1.13
6?1.
056
1.05
8??
1.10
3?1.
057
1.13
9?1.
061
1.06
4??
1.10
8?1.
048
1.11
6?1.
082
1.07
1???
1.09
2?1.
049
1.11
4?1.
077
1.06
4???
1.08
8?
hM
achi
nery
and
Equi
pmen
t
11.
004
0.99
01.
052
1.00
30.
980
1.00
20.
996
1.04
71.
003
0.98
81.
018?
0.97
31.
006
0.97
40.
981
1.01
9?0.
975
1.00
20.
983
0.98
3
21.
002
0.95
21.
010
1.01
40.
994
1.00
40.
962
1.01
71.
005
1.00
41.
044
0.96
10.
980
0.97
80.
990
1.03
60.
955
0.97
70.
984
0.98
8
31.
079
1.08
7?1.
086
1.02
31.
102??
1.07
71.
091?
1.09
01.
033
1.12
6??
1.09
81.
060
1.07
71.
046
1.10
7??
1.09
51.
057
1.08
11.
030
1.09
1?
41.
060
1.05
7?1.
033
0.96
31.
053
1.06
11.
047
1.03
30.
960
1.07
0?1.
099
1.01
71.
048
0.94
61.
048
1.09
21.
023
1.04
70.
928
1.03
4
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
2
Tabl
e3.
A1.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(2
)and
fact
or-a
ugm
ente
dA
R(2
)mod
el(*
)PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEICCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
026
1.08
7???
1.02
61.
011
1.05
2??
1.01
41.
091???
1.03
41.
009
1.05
0?0.
968
1.10
7??
1.06
01.
018
1.05
90.
983
1.09
8?1.
061
1.01
41.
054
20.
980
1.17
3???
1.00
81.
027?
1.09
3??
0.97
21.
171??
1.00
41.
032
1.09
0??
0.93
0??
1.16
8??
1.00
41.
076??
1.10
5?0.
942??
1.16
2??
1.00
21.
067??
1.10
0?
30.
954?
1.20
8??
1.02
61.
024
1.09
9?0.
946?
1.20
6??
1.00
91.
030
1.09
7?0.
894??
1.21
2???
0.97
91.
093?
1.12
1?0.
903??
1.20
4??
0.98
61.
083
1.11
8?
40.
972
1.17
9???
1.03
41.
019
1.07
90.
960??
1.18
0???
1.01
91.
024
1.07
90.
890???
1.19
4???
0.99
21.
082
1.10
60.
906???
1.18
3??
1.00
21.
073
1.10
4h
Dur
able
Con
sum
ptio
n1
1.00
30.
977
0.94
80.
928
0.95
10.
989
0.97
90.
947
0.88
80.
925?
0.98
71.
001
0.94
90.
915
0.95
30.
986
0.99
90.
936
0.90
80.
952
20.
997
0.99
70.
976
0.94
40.
972
0.99
31.
002
0.96
70.
938
0.95
91.
002
1.00
10.
983
0.96
30.
961
0.99
61.
002
0.96
80.
965
0.96
63
1.00
10.
997
0.91
80.
976
0.96
70.
995
0.99
60.
909
0.96
90.
955
1.00
40.
993
0.92
70.
980
0.95
50.
999
0.99
90.
916
0.97
70.
961
41.
005
0.98
30.
909
0.95
50.
929
1.00
00.
987
0.90
00.
946
0.92
11.
008
0.98
80.
904
0.94
50.
920
1.00
40.
991
0.89
60.
941
0.92
6h
Non
-dur
able
Con
sum
ptio
n1
1.02
91.
101???
1.04
3??
1.00
71.
062???
1.02
01.
097???
1.04
3??
1.01
01.
059??
0.97
51.
108??
1.02
51.
028
1.06
10.
985
1.09
9??
1.03
91.
024
1.05
82
0.99
11.
146??
1.08
4?0.
971
1.05
90.
982
1.14
4??
1.06
1?0.
988
1.06
00.
924???
1.16
6??
1.00
61.
049
1.08
20.
936???
1.15
8??
1.02
01.
037
1.08
03
0.97
71.
154??
1.11
8??
0.95
81.
054
0.96
6?1.
153??
1.09
1??
0.97
41.
056
0.89
7???
1.18
2???
1.02
41.
053
1.08
20.
911???
1.17
2???
1.04
01.
041
1.08
24
0.98
81.
134???
1.10
4??
0.94
21.
038
0.97
6?1.
135???
1.07
9??
0.96
11.
041
0.89
6???
1.17
2???
1.02
41.
037
1.07
00.
914???
1.15
9???
1.04
21.
026
1.07
1h
Gro
ssFi
xed
Cap
ital
Form
atio
n1
1.01
70.
992
1.05
80.
985
0.97
71.
025
1.01
21.
049
0.99
60.
996
1.05
71.
010
1.01
10.
947
0.97
11.
053
0.99
61.
005
0.95
30.
967
20.
999
0.99
71.
016
0.92
50.
972
1.00
61.
018
1.01
50.
931
0.99
41.
049
1.01
40.
985
0.87
70.
948?
1.04
10.
996
0.97
90.
872
0.93
2?
31.
007
1.07
21.
007
0.83
20.
983
1.00
81.
090?
0.99
70.
847
1.01
51.
028
1.06
51.
008
0.81
70.
957
1.02
51.
047
1.00
10.
795
0.93
0?
40.
992
1.06
51.
009
0.83
70.
973
0.99
41.
075?
1.00
00.
843
1.00
01.
004
1.04
91.
020
0.82
20.
950
1.00
21.
035
1.00
90.
804
0.92
7?
hC
onst
ruct
ion
and
Wor
ks1
1.03
41.
054
1.07
11.
029
1.03
01.
028
1.05
71.
074
1.03
61.
033
1.02
1??
1.02
7??
1.06
01.
031
1.01
51.
022??
1.02
4??
1.05
81.
030
1.01
32
1.00
31.
048
1.02
30.
998
1.02
11.
002
1.05
01.
033
1.00
71.
024
1.00
31.
028??
1.03
11.
006
1.01
01.
002
1.02
5??
1.02
81.
003
1.00
73
0.99
01.
055?
0.98
60.
979
1.02
40.
993
1.05
8?0.
994
0.98
71.
028
0.99
11.
037??
1.00
50.
990
1.01
30.
991
1.03
4??
1.00
10.
985
1.00
94
0.98
31.
060??
0.98
50.
987
1.02
9?0.
985
1.06
2??
0.98
90.
992
1.03
3??
0.97
7?1.
040??
1.00
90.
999
1.01
80.
978?
1.03
9??
1.00
40.
992
1.01
4h
Mac
hine
ryan
dEq
uipm
ent
11.
030?
1.01
51.
079??
1.02
81.
005
1.02
7?1.
021
1.07
4??
1.02
91.
014
1.04
40.
998
1.03
10.
999
1.00
61.
045?
1.00
01.
028
1.00
81.
008
21.
012
0.96
11.
020
1.02
41.
004
1.01
40.
972
1.02
71.
015
1.01
41.
054
0.97
00.
990
0.98
81.
000
1.04
60.
965
0.98
70.
994
0.99
73
1.02
61.
034
1.03
30.
973
1.04
81.
024
1.03
71.
037?
0.98
21.
071
1.04
41.
008
1.02
40.
995
1.05
31.
042
1.00
51.
028?
0.98
01.
037
41.
007
1.00
50.
982
0.91
5??
1.00
11.
008
0.99
50.
982
0.91
3???
1.01
71.
044
0.96
70.
996
0.89
9??
0.99
61.
038
0.97
30.
995
0.88
2??
0.98
3(*
)Rel
ativ
eR
MSF
E1 h(s
eeeq
uati
on3)
.Ora
nge-
shad
edce
lls=F
igur
esbe
low
1.G
reen
-sha
ded
cells
=Fig
ures
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Cla
rkan
dW
est(
2007
)tes
t’sp-
valu
e:(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
3
Tabl
e4.
A1:
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
fact
or-a
ugm
ente
dA
R(2
)mod
els
acco
rdin
gto
its
rela
tion
ship
wit
hth
ebu
sine
sscy
cle
(*)
PCSI
PFEI
CCSI
CFEI
OFEIPCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
022
1.04
9?0.
960
1.07
31.
081
1.00
01.
070??
0.96
61.
059
1.07
80.
968
1.09
11.
032
1.05
31.
066
0.96
31.
010
1.04
61.
020
1.01
0
20.
953?
1.04
60.
918?
0.91
80.
973
0.94
4??
1.05
2?0.
916?
0.92
20.
973
0.87
5??
1.04
40.
906
0.96
90.
977
0.88
1??
1.00
10.
912
0.95
30.
956
30.
959?
1.05
6?0.
922??
0.89
5??
0.96
00.
953??
1.05
6?0.
907??
0.89
9??
0.95
70.
856???
1.06
30.
868?
0.98
10.
977
0.86
8??
1.02
10.
880?
0.97
60.
969
40.
978
1.02
20.
929??
0.88
7??
0.93
80.
968?
1.02
70.
914??
0.89
3??
0.93
80.
845???
1.04
50.
882?
0.98
50.
969
0.86
5??
1.00
30.
905?
0.98
30.
968
hD
urab
leC
onsu
mpt
ion
11.
028
0.97
10.
992
0.99
40.
987
1.00
20.
973
0.99
10.
916
0.93
90.
973
0.99
90.
985
0.95
60.
978
0.97
20.
975
0.95
70.
928??
0.96
1
20.
999
0.99
51.
008
0.95
60.
999
0.99
01.
003
0.99
30.
938
0.97
50.
992
0.99
71.
005
0.98
80.
973
0.98
70.
987
0.96
91.
006
0.99
2
31.
017
1.01
81.
041
0.99
11.
013
1.00
71.
019
1.02
40.
973
0.98
91.
016
1.01
21.
036
0.99
80.
994
1.00
91.
011
1.01
61.
002
1.02
0
41.
016
1.01
01.
044
1.01
11.
014
1.00
41.
018
1.02
80.
994
0.99
61.
009
1.02
21.
008
0.99
51.
008
1.00
11.
026
0.99
00.
984
1.04
0
hN
on-d
urab
leC
onsu
mpt
ion
10.
992
1.08
3??
0.98
50.
991
1.04
50.
977
1.07
9???
0.99
20.
979
1.03
10.
967
1.08
80.
984
0.98
70.
997
0.95
21.
027
1.03
30.
978
0.96
6
20.
961???
1.07
2??
0.96
60.
869???
0.95
30.
951??
1.06
8??
0.94
50.
886??
0.95
40.
884??
1.10
9???
0.85
20.
970
0.97
80.
888??
1.07
5??
0.88
10.
965
0.97
7
30.
961??
1.07
1???
0.96
0?0.
836??
0.93
80.
951??
1.06
8???
0.93
5?0.
855??
0.93
90.
862???
1.12
2???
0.84
8?0.
971
0.96
90.
875???
1.08
4??
0.87
5?0.
975
0.97
7
40.
977
1.04
70.
957
0.83
2??
0.92
40.
966??
1.04
70.
932
0.85
5??
0.92
70.
859???
1.11
8???
0.86
40.
983
0.96
80.
879???
1.07
8???
0.89
50.
989
0.98
2
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
060
1.00
01.
000
1.01
50.
975
1.06
31.
027
1.00
41.
037
1.00
81.
142
1.00
90.
923
0.94
00.
954
1.12
70.
949
0.89
60.
929
0.91
8
21.
006
0.94
60.
942
1.00
80.
934
1.01
00.
977
0.95
41.
036
0.98
11.
109
0.98
60.
912
0.89
90.
879
1.08
50.
926
0.89
90.
836
0.80
6?
31.
004
0.90
40.
841??
0.92
80.
863?
1.00
50.
930
0.83
1??
0.98
40.
925
1.02
80.
898
0.86
70.
867
0.80
2?1.
026
0.83
70.
864
0.75
4?0.
712??
41.
030
0.90
10.
823???
0.91
70.
857??
1.03
50.
918
0.81
3???
0.95
90.
912
1.01
50.
875
0.86
6?0.
866
0.80
1??
1.01
70.
829
0.85
0??
0.76
70.
721??
hC
onst
ruct
ion
and
Wor
ks
11.
015
0.98
60.
966
0.98
90.
979
1.01
20.
993
0.97
41.
002
0.98
70.
958
0.93
40.
952
0.98
90.
950
0.96
10.
905
0.94
10.
988
0.93
4
20.
991
0.98
60.
948?
0.98
70.
980??
0.99
20.
991
0.96
31.
005
0.98
90.
976
0.94
90.
975
0.99
60.
950
0.97
60.
925
0.96
50.
975
0.92
2
30.
986
0.98
6?0.
927??
0.97
9?0.
981??
0.99
20.
994
0.93
5?0.
995
0.99
20.
979
0.95
10.
973
0.99
20.
945
0.98
10.
925
0.96
10.
963
0.91
6
40.
988
.974??
0.90
8??
0.96
7??
0.97
1?0.
994
0.98
1?0.
911??
0.98
10.
984?
0.96
40.
933
0.96
60.
985
0.93
80.
966
0.90
80.
948?
0.95
80.
911
hM
achi
nery
and
Equi
pmen
t
11.
070
1.07
71.
046
1.03
31.
024
1.05
71.
089
1.04
71.
033
1.03
61.
108
1.02
00.
933
0.99
21.
021
1.11
61.
006
0.90
51.
007
1.02
4
21.
012
1.02
11.
030
1.03
11.
009
1.00
61.
039
1.05
21.
024
1.02
71.
130
1.02
10.
980
0.97
00.
991
1.11
90.
985
0.96
80.
973
0.97
5
31.
017
0.97
10.
944
0.90
40.
928
1.00
10.
976
0.95
70.
928
0.96
11.
085
0.91
60.
943
0.94
40.
922
1.09
80.
870?
0.96
10.
910
0.88
3
41.
036
1.01
80.
911
0.94
00.
956
1.02
71.
012
0.91
20.
951
0.98
31.
137
0.96
70.
932
0.91
90.
935
1.13
60.
943
0.93
30.
862
0.88
8
(?)R
elat
ive
RM
SFE2 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=
Figu
res
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Har
vey,
Leyb
ourn
e,an
dN
ewbo
ld(1
997)
test
’sp-
valu
e:(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
4
2R
obus
tnes
sre
sult
sII
:AR
(3)n
ativ
em
odel
,4-t
erm
fact
orin
diff
eren
ces
Tabl
e1.
A2.
Adj
uste
dR
-squ
ared
rati
obe
twee
nA
R(3
)fac
tor-
augm
ente
dan
dna
tive
AR
(3)m
odel
(*)
and
F-te
stre
sult
s(p
-val
ue)o
fthe
cons
umer
-bas
edfa
ctor
stat
isti
cals
igni
fican
ce(*
*)Fa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sFa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEI
Fact
or:C
omm
erce
+In
dust
ryPC
0.98
31.
017
0.99
40.
999
1.01
10.
985
0.99
81.
017
0.99
71.
000
0.96
80.
121
0.55
90.
298
0.07
90.
961
0.52
10.
024
0.56
30.
358
DC
0.98
71.
033
0.98
71.
002
1.02
10.
984
0.99
81.
002
0.99
00.
996
0.74
00.
002
0.74
90.
314
0.03
30.
968
0.37
10.
091
0.62
20.
411
ND
C0.
947
1.03
20.
954
0.99
11.
023
0.95
40.
990
0.97
90.
980
0.99
50.
980
0.04
40.
900
0.16
60.
033
0.88
60.
405
0.51
20.
380
0.19
1G
FCF
0.98
91.
005
1.01
91.
006
1.01
31.
009
0.99
61.
020
1.01
31.
016
0.83
80.
312
0.01
80.
338
0.14
20.
308
0.44
40.
111
0.28
20.
153
CW
K0.
994
0.99
91.
012
1.02
71.
019
1.00
11.
003
1.00
51.
017
1.01
60.
979
0.87
90.
389
0.06
00.
156
0.66
20.
672
0.48
20.
153
0.12
9M
EQ0.
990
1.00
71.
023
1.00
91.
018
1.02
10.
986
1.07
21.
030
1.03
40.
804
0.33
70.
100
0.50
00.
257
0.18
80.
634
0.02
60.
399
0.24
1Fa
ctor
:Com
mer
ce+
Indu
stry
+Tr
ansp
orta
tion
PC0.
985
1.01
01.
001
0.99
91.
007
0.98
31.
003
1.01
30.
997
1.00
20.
953
0.20
60.
300
0.34
60.
138
0.98
50.
383
0.06
00.
538
0.28
7D
C0.
990
1.03
10.
990
0.99
81.
013
0.98
40.
998
0.99
90.
991
0.99
80.
597
0.00
10.
492
0.43
80.
128
0.97
50.
403
0.16
00.
549
0.30
2N
DC
0.94
91.
024
0.95
60.
985
1.01
10.
952
0.99
40.
978
0.98
51.
002
0.96
30.
136
0.87
70.
295
0.08
30.
875
0.32
40.
535
0.28
90.
126
GFC
F0.
995
1.00
21.
023
1.00
61.
012
0.99
70.
995
1.02
01.
014
1.01
60.
785
0.40
20.
011
0.36
80.
183
0.55
10.
463
0.12
00.
239
0.12
0C
WK
0.99
31.
000
1.01
01.
027
1.02
10.
999
1.00
31.
007
1.01
71.
015
0.98
50.
849
0.46
00.
044
0.07
80.
728
0.72
10.
403
0.19
00.
186
MEQ
1.00
40.
999
1.02
91.
002
1.01
20.
996
0.98
81.
081
1.03
91.
042
0.73
00.
430
0.07
20.
587
0.35
10.
359
0.66
00.
015
0.33
60.
178
Fact
or:C
omm
erce
+In
dust
ry+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
991
0.99
41.
016
0.99
60.
996
0.98
51.
027
0.99
41.
001
1.01
50.
650
0.68
60.
020
0.60
20.
492
0.95
70.
028
0.63
60.
260
0.03
5D
C0.
983
0.99
71.
000
0.99
10.
997
0.98
81.
032
0.98
90.
998
1.01
50.
971
0.41
40.
096
0.63
50.
413
0.73
30.
000
0.65
60.
349
0.05
5N
DC
0.95
90.
978
0.97
40.
982
0.99
20.
954
1.05
70.
952
0.98
61.
023
0.76
40.
592
0.53
40.
347
0.25
10.
955
0.02
00.
953
0.22
70.
021
GFC
F1.
011
0.99
51.
022
1.01
51.
017
0.98
71.
004
1.01
21.
005
1.01
20.
274
0.49
20.
087
0.22
70.
121
0.91
40.
322
0.05
90.
407
0.19
5C
WK
1.00
61.
005
1.00
51.
018
1.01
50.
995
0.99
71.
012
1.02
51.
019
0.36
60.
589
0.52
40.
132
0.14
30.
953
0.93
60.
337
0.07
10.
126
MEQ
1.01
90.
987
1.07
21.
038
1.04
00.
984
1.00
11.
026
1.00
31.
014
0.18
10.
655
0.02
80.
306
0.17
40.
850
0.39
30.
095
0.63
20.
370
Fact
or:C
omm
erce
+C
onst
ruct
ion
+Pe
rson
alSe
rvic
esPC
0.98
71.
002
1.01
80.
999
1.00
30.
986
1.02
20.
997
0.99
51.
005
0.87
30.
299
0.01
40.
422
0.21
40.
969
0.05
50.
597
0.43
50.
166
DC
0.98
31.
006
0.99
90.
997
1.00
60.
990
1.03
30.
992
0.98
91.
001
0.96
30.
122
0.13
80.
381
0.16
50.
714
0.00
00.
529
0.71
80.
349
ND
C0.
954
0.99
40.
968
0.99
31.
010
0.96
11.
045
0.97
10.
969
0.99
60.
888
0.31
40.
585
0.20
80.
092
0.89
60.
080
0.65
70.
533
0.15
9G
FCF
1.00
51.
001
1.02
51.
014
1.01
90.
990
0.99
31.
000
1.00
11.
001
0.45
30.
429
0.05
90.
237
0.10
40.
880
0.67
50.
319
0.49
20.
407
CW
K1.
001
1.00
41.
006
1.02
01.
017
0.99
60.
996
1.01
11.
024
1.01
90.
633
0.64
80.
458
0.10
70.
129
0.95
10.
923
0.38
50.
061
0.09
9M
EQ1.
014
1.00
21.
072
1.03
91.
047
0.99
30.
975
1.01
30.
986
0.98
80.
299
0.52
60.
023
0.27
50.
132
0.74
20.
791
0.17
60.
819
0.70
3(*
)Adj
uste
dR
-squ
ared
rati
o=A
djus
ted
R-s
quar
edw
ith
the
fact
or/A
djus
ted
R-s
quar
edw
itho
utth
efa
ctor
.Ora
nge-
shad
edce
lls=A
djus
ted
R-s
quar
edra
tio
grea
ter
than
1.G
reen
-sha
ded
cells
=Adj
uste
dR
-squ
ared
rati
ogr
eate
rth
an1.
05(5
%of
adju
stm
entg
ain)
.Sam
ple:
2001
.I-20
14.IV
(56
obse
rvat
ions
).(*
*)p-
valu
eof
the
null
hypo
thes
is:N
H:φ
1=...
=φ
4=0
(see
equa
tion
1)ag
ains
tthe
alte
rnat
ive
ofat
leas
tone
term
iseq
ual
toze
ro.G
reen
-sha
ded
cells
=p<
10%
.Ora
nge-
shad
edce
lls=p<
15%
.Sam
ple:
2001
.I-20
19.II
I(75
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
5
Tabl
e2.
A2.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(3
)mod
el(*
)PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
339??
1.48
3???
1.34
6??
1.30
6??
1.39
6??
1.32
0??
1.49
0???
1.33
9??
1.31
2??
1.40
0??
1.25
9??
1.48
6???
1.39
0??
1.34
5??
1.40
6??
1.28
1??
1.47
6???
1.38
3??
1.33
5??
1.40
1??
21.
658???
1.97
3???
1.69
0??
1.74
1???
1.84
4???
1.64
3???
1.98
3???
1.67
8??
1.75
1??
1.84
4???
1.57
7???
1.96
6???
1.70
4??
1.80
4???
1.85
5???
1.59
6???
1.94
7???
1.69
3??
1.79
3???
1.84
8???
31.
983???
2.38
9???
2.09
8???
2.10
3???
2.22
2???
1.97
0???
2.40
6???
2.07
3???
2.10
9???
2.22
3???
1.88
4???
2.40
0???
2.05
7??
2.19
1???
2.24
8???
1.90
0???
2.37
4???
2.05
6??
2.17
8???
2.24
2???
42.
176???
2.52
9???
2.27
6???
2.26
2???
2.37
0???
2.15
7???
2.54
5???
2.25
6???
2.27
1???
2.37
2???
2.03
9???
2.55
1???
2.25
2???
2.35
2???
2.40
3???
2.06
6???
2.52
2???
2.25
4???
2.33
9???
2.39
8???
hD
urab
leC
onsu
mpt
ion
10.
995
0.99
20.
951
0.90
80.
963
0.98
70.
996
0.94
20.
867
0.94
00.
982
1.00
00.
962
0.91
50.
964
0.98
00.
999
0.94
70.
902
0.96
1
20.
937
0.94
90.
917
0.91
60.
949
0.93
60.
957
0.90
60.
909
0.93
90.
943
0.95
10.
929
0.93
90.
941
0.93
80.
953
0.91
40.
937
0.94
3
30.
904
0.92
50.
836
0.90
60.
915
0.90
10.
928
0.82
70.
899
0.90
40.
908
0.91
70.
852
0.91
10.
901
0.90
40.
924
0.84
30.
906
0.90
4
40.
856
0.86
40.
781
0.85
20.
837
0.85
30.
870
0.77
20.
844
0.83
10.
859
0.85
90.
787
0.84
50.
825
0.85
60.
861
0.78
10.
839
0.82
7
hN
on-d
urab
leC
onsu
mpt
ion
11.
341??
1.47
0???
1.36
1???
1.28
9??
1.39
3???
1.32
6??
1.47
5???
1.34
7???
1.30
1??
1.39
4??
1.24
4??
1.47
6???
1.35
0??
1.33
6??
1.39
3??
1.26
3??
1.46
0???
1.35
7??
1.32
9??
1.39
0??
21.
615???
1.85
8???
1.72
8???
1.60
9???
1.73
9???
1.59
7???
1.86
8???
1.69
7???
1.62
2??
1.73
7???
1.49
1???
1.88
9???
1.66
6???
1.69
4???
1.75
7???
1.51
6???
1.86
7???
1.67
2???
1.67
8???
1.75
3???
31.
858???
2.14
1???
2.05
3???
1.85
6???
2.00
7???
1.84
0???
2.15
7???
2.01
7???
1.86
7???
2.00
5???
1.71
9???
2.18
5???
1.96
9???
1.96
8???
2.03
1???
1.74
4???
2.15
8???
1.97
4???
1.95
0???
2.02
9???
42.
038???
2.29
7???
2.19
7???
2.00
4???
2.15
8???
2.01
9???
2.31
2???
2.16
6???
2.01
9???
2.15
8???
1.88
3???
2.35
7???
2.13
3???
2.12
0???
2.19
1???
1.91
5???
2.32
4???
2.14
0???
2.10
3???
2.18
9???
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
108
1.09
0?1.
166??
1.05
6?1.
058
1.11
4?1.
109??
1.15
6??
1.07
01.
078
1.13
1?1.
093?
1.11
71.
016
1.04
61.
133?
1.07
91.
108
1.02
11.
040
21.
021
1.03
2?1.
059?
0.93
10.
979
1.02
71.
054??
1.05
7?0.
940
1.00
31.
051
1.03
41.
017
0.88
30.
949
1.05
01.
014
1.01
20.
880
0.93
4
31.
069
1.12
5?1.
066
0.94
51.
035
1.07
41.
146??
1.05
60.
952
1.06
31.
087
1.11
6?1.
045
0.92
01.
008
1.08
81.
096
1.04
40.
908
0.98
7
41.
109?
1.16
4?1.
120?
1.02
5?1.
087
1.11
6?1.
176?
1.11
2?1.
021?
1.10
81.
123
1.14
9?1.
108
1.00
2?1.
064
1.12
5?1.
135?
1.10
50.
996?
1.05
0
hC
onst
ruct
ion
and
Wor
ks
10.
957
0.97
80.
992
0.95
30.
954
0.95
10.
980
0.99
70.
957
0.95
70.
947
0.95
40.
987
0.95
20.
942
0.94
70.
953
0.98
40.
951
0.94
0
21.
029
1.07
4?1.
049?
1.02
9?1.
046
1.02
81.
075?
1.06
1?1.
036?
1.04
91.
031
1.05
31.
059?
1.03
3?1.
036
1.03
01.
051
1.05
6?1.
030?
1.03
3
31.
108?
1.17
4??
1.10
6??
1.10
4??
1.14
0??
1.11
0?1.
177??
1.11
5??
1.11
0??
1.14
6??
1.11
0?1.
153??
1.12
8??
1.10
9??
1.13
0??
1.10
9?1.
151??
1.12
5??
1.10
5??
1.12
5??
41.
163??
1.24
3??
1.16
7??
1.17
7???
1.20
7??
1.16
4??
1.24
4??
1.17
4??
1.18
0???
1.21
2??
1.15
9??
1.21
8??
1.19
7??
1.18
2???
1.19
7??
1.15
9??
1.21
7??
1.19
2??
1.17
7???
1.19
3??
hM
achi
nery
and
Equi
pmen
t
11.
055?
1.00
71.
145??
1.03
41.
024
1.04
6?1.
000
1.14
5??
1.03
21.
029
1.05
90.
982
1.08
21.
014
1.01
81.
057
0.98
81.
080
1.01
91.
019
21.
058
1.01
21.
140??
1.03
31.
075??
1.05
31.
011
1.15
2??
1.02
81.
087??
1.06
4?0.
990
1.10
4?1.
019
1.06
0??
1.06
2?0.
990
1.10
1?1.
015
1.05
2?
31.
163
1.12
91.
241??
1.12
91.
260??
1.16
11.
124
1.26
1??
1.13
61.
277??
1.17
01.
101?
1.21
4?1.
166?
1.24
6??
1.17
1?1.
102?
1.22
4?1.
140
1.22
8??
41.
047
1.02
51.
120
0.99
41.
133
1.05
21.
016
1.13
30.
988
1.15
21.
085
1.01
21.
111
0.99
41.
119
1.07
71.
014
1.12
00.
968
1.09
8
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
6
Tabl
e3.
A2.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(3
)and
fact
or-a
ugm
ente
dA
R(3
)mod
el(*
)PCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEI
OFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
022
1.13
2??
1.02
70.
997
1.06
5?1.
007
1.13
7???
1.02
21.
001
1.06
8?0.
961
1.13
4??
1.06
11.
026
1.07
30.
978
1.12
6??
1.05
61.
019
1.06
92
0.98
81.
175???
1.00
71.
037
1.09
9??
0.97
91.
181???
1.00
01.
043
1.09
9??
0.93
9??
1.17
1???
1.01
51.
075?
1.10
5??
0.95
1??
1.16
0??
1.00
91.
068?
1.10
1??
30.
975
1.17
4??
1.03
11.
033
1.09
2??
0.96
81.
182???
1.01
91.
036
1.09
3?0.
926??
1.17
9???
1.01
11.
077
1.10
5?0.
934??
1.16
7??
1.01
01.
071
1.10
2?
40.
984
1.14
4???
1.03
01.
023
1.07
2?0.
975?
1.15
1???
1.02
11.
027
1.07
3?0.
922???
1.15
4???
1.01
91.
064
1.08
7?0.
935???
1.14
1???
1.01
91.
058
1.08
5?
hD
urab
leC
onsu
mpt
ion
11.
001
0.99
80.
957
0.91
30.
969
0.99
31.
002
0.94
80.
872?
0.94
60.
988
1.00
60.
968
0.92
00.
970
0.98
61.
005
0.95
30.
908
0.96
72
0.99
51.
008
0.97
40.
973
1.00
90.
994
1.01
70.
963
0.96
60.
998
1.00
21.
010
0.98
70.
997
1.00
00.
997
1.01
30.
971
0.99
51.
002
31.
001
1.02
50.
926
1.00
41.
013
0.99
81.
028
0.91
60.
995
1.00
11.
006
1.01
50.
944
1.00
90.
997
1.00
11.
023
0.93
41.
003
1.00
14
1.00
91.
019
0.92
11.
005
0.98
71.
006
1.02
60.
911
0.99
60.
980
1.01
31.
013
0.92
80.
997
0.97
31.
010
1.01
60.
921
0.98
90.
976
hN
on-d
urab
leC
onsu
mpt
ion
11.
031
1.13
1??
1.04
7?0.
992
1.07
2?1.
020
1.13
5???
1.03
61.
001
1.07
3?0.
957
1.13
6??
1.03
91.
028
1.07
20.
971
1.12
3??
1.04
41.
022
1.06
92
0.99
91.
150??
1.06
9?0.
995
1.07
6?0.
988
1.15
6??
1.05
01.
003
1.07
50.
922???
1.16
9???
1.03
11.
048
1.08
70.
938???
1.15
5??
1.03
41.
038
1.08
4?
30.
989
1.14
0???
1.09
3??
0.98
81.
068
0.97
91.
148???
1.07
3?0.
994
1.06
70.
915???
1.16
3???
1.04
81.
048
1.08
10.
928???
1.14
9???
1.05
11.
038
1.08
04
0.99
31.
119???
1.07
1??
0.97
71.
052
0.98
41.
126???
1.05
6?0.
984
1.05
20.
918???
1.14
9???
1.04
01.
033
1.06
80.
933???
1.13
3???
1.04
31.
025
1.06
7h
Gro
ssFi
xed
Cap
ital
Form
atio
n1
1.02
51.
008
1.07
8?0.
976
0.97
81.
030?
1.02
51.
069??
0.99
00.
996
1.04
61.
010
1.03
30.
940
0.96
71.
048?
0.99
71.
025
0.94
40.
962
21.
006
1.01
61.
043
0.91
70.
964
1.01
21.
038
1.04
10.
926
0.98
81.
035
1.01
81.
002
0.87
00.
935?
1.03
40.
998
0.99
70.
867
0.92
0?
30.
999
1.05
1?0.
996
0.88
30.
967
1.00
41.
071?
0.98
70.
890
0.99
31.
016
1.04
30.
977
0.85
90.
942
1.01
71.
025
0.97
60.
849
0.92
24
0.99
01.
039?
1.00
00.
915
0.97
00.
995
1.04
9?0.
993
0.91
10.
988
1.00
21.
025
0.98
90.
894
0.94
91.
003
1.01
20.
986
0.88
90.
937
hC
onst
ruct
ion
and
Wor
ks1
1.02
91.
053
1.06
7?1.
025
1.02
61.
023
1.05
41.
072?
1.03
01.
029
1.01
81.
027??
1.06
21.
025
1.01
31.
018
1.02
5??
1.05
81.
024
1.01
12
1.00
21.
046
1.02
21.
002
1.01
81.
001
1.04
71.
033
1.00
91.
022
1.00
41.
026?
1.03
11.
006
1.00
91.
003
1.02
4?1.
028
1.00
31.
006
30.
992
1.05
20.
991
0.98
91.
022
0.99
51.
054
0.99
90.
995
1.02
60.
994
1.03
3??
1.01
00.
993
1.01
20.
993
1.03
1??
1.00
70.
990
1.00
84
0.98
71.
055?
0.99
00.
999
1.02
5??
0.98
81.
056?
0.99
61.
002
1.02
9??
0.98
4?1.
034??
1.01
61.
003
1.01
6?0.
984?
1.03
2??
1.01
20.
999
1.01
2h
Mac
hine
ryan
dEq
uipm
ent
11.
029??
0.98
21.
116???
1.00
80.
998
1.01
9?0.
975
1.11
6???
1.00
61.
003
1.03
20.
957
1.05
50.
988
0.99
21.
030
0.96
31.
052?
0.99
30.
993
21.
020
0.97
51.
099???
0.99
61.
036
1.01
50.
974
1.11
1???
0.99
11.
048
1.02
60.
954
1.06
40.
983
1.02
21.
024
0.95
41.
061?
0.97
91.
014
31.
017
0.98
71.
086??
0.98
71.
102
1.01
60.
984
1.10
3???
0.99
41.
117
1.02
30.
963
1.06
2??
1.02
01.
090
1.02
40.
964
1.07
1??
0.99
71.
074
40.
994
0.97
31.
063?
0.94
31.
075
0.99
90.
964
1.07
5??
0.93
8?1.
093
1.03
00.
960
1.05
5??
0.94
31.
062
1.02
20.
962
1.06
3??
0.91
8??
1.04
2(*
)Rel
ativ
eR
MSF
E1 h(s
eeeq
uati
on3)
.Ora
nge-
shad
edce
lls=F
igur
esbe
low
1.G
reen
-sha
ded
cells
=Fig
ures
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Cla
rkan
dW
est(
2007
)tes
t’sp-
valu
e:(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
7
Tabl
e4.
A2:
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
fact
or-a
ugm
ente
dA
R(3
)mod
els
acco
rdin
gto
its
rela
tion
ship
wit
hth
ebu
sine
sscy
cle
(*)
PCSI
PFEICCSI
CFEIOFEIPCSI
PFEICCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEICCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
977
1.02
80.
907
0.96
91.
012
0.95
51.
047
0.90
0?0.
978
1.02
20.
911
1.02
50.
961
1.01
51.
007
0.91
80.
968
0.96
00.
990
0.96
9
20.
953??
1.02
30.
894
0.91
9?0.
963
0.94
4??
1.03
60.
889?
0.93
00.
968
0.87
5??
1.01
80.
907
0.95
90.
962
0.88
3?0.
974
0.90
00.
946
0.94
0
30.
972
1.03
10.
924
0.92
1?0.
966
0.96
91.
043
0.91
2?0.
926?
0.96
80.
885?
1.03
80.
906?
0.97
70.
973
0.89
4?0.
992
0.90
20.
972
0.95
8
40.
997
1.00
80.
935?
0.91
2?0.
950
0.99
11.
019
0.92
7?0.
920?
0.95
20.
888??
1.02
70.
928
0.97
80.
967
0.90
2??
0.98
80.
934
0.97
70.
960
hD
urab
leC
onsu
mpt
ion
11.
015
0.96
70.
926
0.97
00.
974
0.99
60.
970
0.91
10.
896?
0.93
1?0.
961
0.95
20.
932
0.99
40.
961
0.96
80.
943
0.90
60.
947
0.95
0
20.
978
0.96
50.
932
0.95
80.
984
0.97
40.
976
0.91
30.
943?
0.96
70.
969
0.95
10.
959
0.99
80.
955
0.97
20.
947
0.91
9?0.
996
0.96
4
31.
005
0.99
00.
976
0.99
31.
002
1.00
10.
994
0.95
80.
975
0.98
11.
004
0.97
21.
007
0.99
70.
973
1.00
80.
974
0.98
80.
990
0.99
0
40.
998
0.97
10.
955
0.98
80.
982
0.99
10.
983
0.94
00.
976
0.97
10.
985
0.96
00.
965
0.97
40.
966
0.98
70.
963
0.95
30.
952
0.98
0
hN
on-d
urab
leC
onsu
mpt
ion
10.
966
1.07
40.
953
0.93
31.
014
0.95
11.
086?
0.94
50.
943
1.01
50.
910
1.07
80.
955
0.96
70.
984
0.90
81.
020
0.98
10.
958
0.95
7
20.
971??
1.06
90.
964
0.90
3??
0.97
80.
959???
1.07
9?0.
947
0.91
2?0.
978
0.88
1??
1.11
0???
0.90
80.
963
0.98
60.
890??
1.06
40.
921
0.95
10.
974
30.
977
1.07
0?0.
969
0.89
3?0.
976
0.96
81.
079??
0.95
10.
899?
0.97
40.
883??
1.11
2??
0.91
7?0.
974
0.98
50.
894??
1.06
60.
925?
0.96
70.
978
40.
999
1.04
90.
970
0.89
4??
0.96
50.
991
1.05
70.
954
0.90
3??
0.96
30.
897???
1.10
6??
0.93
30.
982
0.98
40.
912??
1.06
4?0.
943
0.97
90.
984
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
079
0.98
00.
980
0.99
60.
957
1.08
11.
006
0.98
21.
023
0.99
11.
098
0.97
80.
913
0.92
80.
930
1.09
30.
915
0.88
70.
914
0.89
4
21.
008
0.91
30.
930
0.99
80.
906
1.01
10.
954
0.93
91.
040
0.96
21.
058
0.92
90.
900
0.88
30.
837?
1.04
30.
842
0.88
10.
813
0.75
4??
30.
986
0.84
10.
837??
0.95
20.
844??
0.99
10.
878
0.82
7??
1.00
00.
904??
0.97
90.
840
0.84
9?0.
874
0.78
0??
0.98
40.
757?
0.85
00.
781
0.69
3??
41.
015
0.84
50.
826??
0.95
00.
848??
1.02
20.
870
0.81
9??
0.97
60.
896?
0.97
40.
827
0.85
5??
0.89
40.
794??
0.97
80.
754?
0.84
7??
0.82
30.
718??
hC
onst
ruct
ion
and
Wor
ks
11.
000
0.95
20.
912??
0.97
90.
957?
0.99
60.
956?
0.92
3?0.
992
0.96
8?0.
937
0.89
00.
908?
0.98
00.
925
0.93
60.
861
0.89
10.
976
0.90
9
20.
982
0.95
1??
0.91
2??
0.97
60.
955??
0.98
20.
954??
0.93
1?0.
995
0.96
7??
0.95
90.
905
0.94
30.
981
0.92
3?0.
957
0.88
00.
929
0.96
2??
0.89
7
30.
981
0.94
5??
0.89
5???
0.96
8?0.
949?
0.98
50.
950??
0.90
5??
0.98
50.
965?
0.96
10.
897
0.93
8?0.
974
0.91
40.
962
0.87
30.
929?
0.95
00.
887
40.
981
0.92
8?0.
872??
0.95
5?0.
935?
0.98
50.
933?
0.87
8??
0.97
0??
0.95
1?0.
944
0.87
40.
930?
0.96
7?0.
903
0.94
70.
852
0.91
7??
0.94
50.
879
hM
achi
nery
and
Equi
pmen
t
11.
054
1.04
31.
079
1.00
31.
011
1.03
11.
035
1.08
31.
000
1.01
71.
071
0.98
90.
967
0.98
01.
002
1.07
20.
978
0.94
20.
994
1.00
1
21.
024
1.01
61.
054
0.95
81.
001
1.00
71.
018
1.08
10.
959
1.01
71.
073
0.99
01.
005
0.94
30.
978
1.06
70.
965
1.00
40.
935
0.95
8
31.
038
0.97
40.
973
0.88
60.
972
1.02
50.
971
1.01
00.
898
0.99
31.
077
0.93
90.
951
0.95
10.
954
1.08
00.
894??
0.97
10.
906
0.90
8
41.
001
0.94
30.
885
0.90
60.
945
0.99
80.
939
0.90
70.
916
0.97
71.
069
0.93
00.
890?
0.92
30.
913
1.05
80.
903
0.90
1?0.
833
0.84
9
(?)R
elat
ive
RM
SFE2 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=
Figu
res
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Har
vey,
Leyb
ourn
e,an
dN
ewbo
ld(1
997)
test
’sp-
valu
e:(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
8
3R
obus
tnes
sre
sult
sII
I:A
R(4
)nat
ive
mod
el,4
-ter
mfa
ctor
indi
ffer
ence
s
Tabl
e1.
A3.
Adj
uste
dR
-squ
ared
rati
obe
twee
nA
R(4
)fac
tor-
augm
ente
dan
dna
tive
AR
(4)m
odel
(*)
and
F-te
stre
sult
s(p
-val
ue)o
fthe
cons
umer
-bas
edfa
ctor
stat
isti
cals
igni
fican
ce(*
*)Fa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sFa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEI
Fact
or:C
omm
erce
+In
dust
ryPC
0.98
51.
008
0.99
10.
992
1.00
10.
986
0.99
11.
005
0.99
00.
992
0.96
60.
288
0.75
50.
625
0.34
90.
942
0.82
10.
159
0.72
10.
604
DC
0.98
61.
033
0.98
61.
002
1.02
10.
983
0.99
71.
002
0.98
90.
995
0.75
90.
002
0.76
30.
296
0.03
90.
968
0.37
50.
067
0.61
50.
422
ND
C0.
971
1.00
40.
975
0.98
81.
001
0.97
70.
973
0.96
50.
976
0.97
70.
823
0.15
00.
787
0.40
60.
174
0.70
20.
711
0.93
30.
630
0.58
1G
FCF
0.98
80.
997
1.00
51.
003
1.00
51.
008
0.99
41.
009
1.00
81.
011
0.87
00.
514
0.06
40.
319
0.17
80.
376
0.52
80.
119
0.21
50.
121
CW
K0.
993
0.99
81.
012
1.02
71.
018
1.00
01.
003
1.00
51.
017
1.01
50.
980
0.87
40.
367
0.06
00.
152
0.65
90.
684
0.42
20.
153
0.13
2M
EQ0.
985
0.97
91.
004
1.00
61.
001
1.02
70.
988
1.04
01.
021
1.02
70.
773
0.62
50.
235
0.38
30.
334
0.11
40.
474
0.04
20.
224
0.14
1Fa
ctor
:Com
mer
ce+
Indu
stry
+Tr
ansp
orta
tion
PC0.
986
1.00
30.
996
0.99
20.
998
0.98
50.
994
1.00
00.
990
0.99
30.
949
0.48
40.
505
0.61
20.
425
0.96
80.
696
0.27
80.
723
0.54
9D
C0.
990
1.03
10.
990
0.99
71.
012
0.98
30.
997
0.99
90.
990
0.99
70.
632
0.00
10.
476
0.41
10.
130
0.97
90.
412
0.13
40.
550
0.32
9N
DC
0.97
00.
991
0.97
50.
985
0.99
40.
978
0.97
70.
963
0.97
70.
981
0.84
30.
409
0.80
60.
491
0.29
80.
745
0.58
40.
956
0.59
20.
465
GFC
F0.
993
0.99
41.
009
1.00
31.
005
1.00
00.
992
1.01
01.
009
1.01
20.
800
0.62
40.
037
0.36
20.
211
0.54
70.
591
0.13
10.
182
0.09
5C
WK
0.99
31.
000
1.01
01.
027
1.02
10.
998
1.00
21.
007
1.01
71.
015
0.98
50.
840
0.42
80.
044
0.07
60.
725
0.73
50.
354
0.19
10.
191
MEQ
0.99
70.
976
1.00
31.
003
1.00
21.
012
0.98
41.
050
1.02
51.
031
0.67
40.
711
0.21
10.
434
0.36
70.
159
0.63
30.
022
0.20
20.
123
Fact
or:C
omm
erce
+In
dust
ry+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
988
0.98
91.
004
0.98
80.
989
0.98
71.
014
0.99
10.
994
1.00
40.
770
0.88
00.
150
0.76
70.
736
0.94
30.
125
0.80
60.
557
0.21
2D
C0.
982
0.99
61.
000
0.99
00.
996
0.98
81.
032
0.98
80.
998
1.01
50.
972
0.41
70.
070
0.62
80.
425
0.74
20.
000
0.67
10.
327
0.05
9N
DC
0.97
10.
968
0.96
50.
976
0.97
60.
984
1.01
80.
973
0.98
61.
001
0.65
40.
808
0.92
90.
622
0.65
50.
614
0.10
20.
831
0.44
90.
140
GFC
F1.
006
0.99
41.
010
1.01
01.
012
0.98
80.
997
1.00
11.
003
1.00
50.
348
0.56
00.
104
0.17
70.
100
0.90
70.
519
0.14
50.
380
0.21
5C
WK
1.00
61.
004
1.00
41.
017
1.01
50.
994
0.99
61.
012
1.02
41.
019
0.36
50.
596
0.48
30.
133
0.14
60.
958
0.93
40.
317
0.07
10.
125
MEQ
1.01
40.
987
1.03
71.
027
1.03
10.
989
0.98
01.
013
1.00
21.
002
0.18
30.
483
0.05
10.
152
0.10
30.
732
0.65
20.
184
0.52
50.
396
Fact
or:C
omm
erce
+C
onst
ruct
ion
+Pe
rson
alSe
rvic
esPC
0.98
60.
994
1.00
50.
991
0.99
40.
988
1.01
40.
994
0.99
00.
997
0.89
90.
669
0.11
00.
637
0.50
20.
941
0.18
60.
705
0.67
70.
456
DC
0.98
21.
005
0.99
90.
996
1.00
50.
990
1.03
30.
991
0.98
81.
001
0.96
40.
123
0.11
10.
384
0.19
30.
723
0.00
00.
565
0.69
40.
349
ND
C0.
967
0.97
00.
962
0.98
10.
983
1.00
31.
023
0.98
60.
978
0.99
20.
832
0.73
90.
961
0.51
10.
452
0.29
60.
083
0.56
80.
597
0.27
1G
FCF
1.00
30.
996
1.01
21.
010
1.01
30.
990
0.99
10.
994
0.99
80.
997
0.46
00.
512
0.07
00.
181
0.09
40.
900
0.73
90.
464
0.51
10.
477
CW
K1.
001
1.00
41.
006
1.02
01.
016
0.99
50.
995
1.01
11.
024
1.01
90.
627
0.65
40.
415
0.10
80.
132
0.95
30.
923
0.35
00.
062
0.10
1M
EQ1.
015
0.98
91.
034
1.02
91.
032
0.99
40.
972
1.01
30.
987
0.98
50.
229
0.53
40.
058
0.13
80.
094
0.66
90.
837
0.18
90.
761
0.72
1(*
)Adj
uste
dR
-squ
ared
rati
o=A
djus
ted
R-s
quar
edw
ith
the
fact
or/A
djus
ted
R-s
quar
edw
itho
utth
efa
ctor
.Ora
nge-
shad
edce
lls=A
djus
ted
R-s
quar
edra
tio
grea
ter
than
1.G
reen
-sha
ded
cells
=Adj
uste
dR
-squ
ared
rati
ogr
eate
rth
an1.
05(5
%of
adju
stm
entg
ain)
.Sam
ple:
2001
.I-20
14.IV
(56
obse
rvat
ions
).(*
*)p-
valu
eof
the
null
hypo
thes
is:N
H:φ
1=...
=φ
4=0
(see
equa
tion
1)ag
ains
tthe
alte
rnat
ive
ofat
leas
tone
term
iseq
ual
toze
ro.G
reen
-sha
ded
cells
=p<
10%
.Ora
nge-
shad
edce
lls=p<
15%
.Sam
ple:
2001
.I-20
19.II
I(75
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
9
Tabl
e2.
A3.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(4
)mod
el(*
)PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
439???
1.72
3???
1.50
1??
1.49
1??
1.58
2???
1.43
0???
1.72
4???
1.50
3??
1.50
1??
1.58
3??
1.39
6???
1.72
6???
1.53
5??
1.56
0??
1.61
8??
1.40
1???
1.72
0???
1.52
0??
1.54
9??
1.61
1??
21.
932???
2.46
2???
2.02
2??
2.10
0???
2.22
8???
1.92
6???
2.46
2???
2.01
8??
2.11
4???
2.22
9???
1.88
7???
2.44
3???
2.03
4??
2.20
0???
2.27
0???
1.89
2???
2.43
5???
2.01
8??
2.18
3???
2.26
0???
32.
412???
3.00
6???
2.54
5???
2.61
4???
2.75
1???
2.40
2???
3.00
8???
2.52
4???
2.62
7???
2.75
2???
2.33
1???
3.01
0???
2.50
5??
2.73
0???
2.80
3???
2.34
2???
2.99
7???
2.50
1???
2.71
4???
2.79
5???
42.
756???
3.22
9???
2.85
9???
2.88
6???
3.00
8???
2.73
9???
3.23
1???
2.84
5???
2.90
2???
3.01
2???
2.63
8???
3.25
8???
2.84
0???
2.99
7???
3.06
5???
2.65
8???
3.24
1???
2.84
0???
2.98
3???
3.05
8???
hD
urab
leC
onsu
mpt
ion
11.
021
1.03
5?0.
998
0.94
81.
003
1.01
51.
033?
0.99
20.
909
0.97
91.
021
1.03
30.
993
0.95
21.
000
1.01
61.
033
0.96
80.
940
0.99
6
20.
937
0.96
70.
954
0.93
30.
969
0.93
90.
970
0.95
1?0.
929
0.96
10.
959
0.95
90.
962
0.95
80.
960
0.95
10.
962
0.94
00.
953
0.95
9
30.
883
0.91
30.
830
0.92
20.
925
0.88
10.
912
0.83
00.
916
0.91
70.
899
0.91
10.
862
0.93
00.
913
0.89
30.
917
0.84
70.
921
0.91
3
40.
848
0.86
50.
776
0.87
00.
856
0.84
50.
868
0.77
40.
865
0.85
20.
855
0.86
90.
799
0.86
90.
845
0.85
20.
871
0.79
20.
857
0.84
4
hN
on-d
urab
leC
onsu
mpt
ion
11.
453???
1.68
7???
1.53
8???
1.47
4???
1.58
0???
1.44
0???
1.68
8???
1.52
7???
1.48
3???
1.57
8???
1.36
1???
1.70
4???
1.51
2???
1.54
4???
1.59
9???
1.37
3???
1.69
3???
1.51
6???
1.53
2???
1.59
5???
21.
873???
2.26
2???
2.05
5???
1.92
8???
2.07
8???
1.85
6???
2.26
7???
2.03
0???
1.94
2???
2.07
7???
1.74
9???
2.29
8???
1.98
9???
2.03
3???
2.11
6???
1.76
9???
2.28
2???
1.98
8???
2.01
3???
2.10
8???
32.
181???
2.58
4???
2.39
8???
2.23
8???
2.39
6???
2.16
4???
2.59
1???
2.36
6???
2.24
6???
2.39
4???
2.04
3???
2.64
2???
2.31
2???
2.35
5???
2.43
6???
2.06
7???
2.62
2???
2.31
3???
2.33
6???
2.43
0???
42.
515???
2.83
9???
2.66
8???
2.51
4???
2.67
0???
2.49
8???
2.84
8???
2.64
3???
2.53
0???
2.67
2???
2.37
1???
2.92
9???
2.60
8???
2.63
1???
2.71
9???
2.39
8???
2.90
1???
2.61
1???
2.61
4???
2.71
4???
hG
ross
Fixe
dC
apit
alFo
rmat
ion
10.
943
1.00
4?0.
962
0.92
60.
955
0.94
51.
000?
0.97
40.
939
0.96
30.
976
0.98
20.
975
0.89
80.
943
0.97
30.
975
0.97
30.
892
0.93
4
21.
049
1.17
2??
1.05
01.
074
1.11
91.
057
1.18
0??
1.07
21.
098
1.13
31.
119
1.16
11.
107
1.06
01.
112
1.11
01.
146
1.09
51.
049
1.10
0
31.
277
1.46
6??
1.28
2?1.
314?
1.37
71.
281
1.46
8??
1.28
71.
339?
1.39
11.
323
1.43
0?1.
325
1.31
5?1.
370
1.31
71.
419?
1.32
41.
301?
1.35
8
41.
476?
1.67
2???
1.53
9??
1.51
2??
1.57
2?1.
478?
1.66
6???
1.54
2?1.
533??
1.58
6?1.
501?
1.61
4??
1.57
8?1.
505??
1.56
0?1.
499?
1.60
4??
1.57
2?1.
492??
1.54
6?
hC
onst
ruct
ion
and
Wor
ks
11.
043
1.09
41.
097?
1.03
31.
045
1.03
61.
088
1.10
5?1.
042
1.04
91.
032
1.04
51.
098
1.03
81.
030
1.03
21.
045
1.09
01.
035
1.02
7
21.
142?
1.21
8??
1.16
4??
1.14
4??
1.16
9?1.
142?
1.21
3?1.
181??
1.15
5??
1.17
3?1.
148?
1.17
6?1.
183?
1.15
3??
1.15
7?1.
146?
1.17
5?1.
178?
1.14
8??
1.15
3?
31.
207?
1.30
2??
1.20
2??
1.21
2??
1.25
4??
1.21
0?1.
300??
1.21
6??
1.22
2??
1.26
0??
1.21
1?1.
267??
1.23
3??
1.22
4??
1.24
3??
1.21
0?1.
266?
1.22
9??
1.21
8??
1.23
9??
41.
284??
1.39
3??
1.30
5??
1.30
3???
1.34
0??
1.28
4??
1.38
9??
1.31
3??
1.31
0???
1.34
5??
1.27
8??
1.35
1??
1.33
8??
1.31
7???
1.33
0??
1.27
8??
1.35
1??
1.33
1??
1.31
0???
1.32
6??
hM
achi
nery
and
Equi
pmen
t
10.
833
0.83
4?0.
903?
0.86
70.
839
0.82
10.
810
0.90
4?0.
860
0.82
90.
830
0.78
20.
882?
0.84
70.
830
0.82
80.
795
0.88
1?0.
854
0.83
9
20.
953
0.95
1?0.
986
1.02
70.
988
0.94
50.
939?
0.99
41.
022
0.98
40.
986
0.92
01.
002
1.00
10.
983
0.98
00.
928
0.99
01.
011
0.99
1
31.
306?
1.30
9??
1.31
2?1.
363?
1.37
2??
1.29
7?1.
296??
1.31
8?1.
357?
1.36
5?1.
339?
1.26
5?1.
325?
1.35
5?1.
357?
1.33
4?1.
272?
1.32
1?1.
354?
1.36
1?
41.
173
1.18
5??
1.18
3?1.
234
1.24
5?1.
172
1.18
1??
1.18
7?1.
235
1.25
0?1.
222
1.19
9??
1.20
3?1.
220
1.24
7?1.
214
1.19
2??
1.20
2?1.
214
1.24
3?
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
10
Tabl
e3.
A3.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(4
)and
fact
or-a
ugm
ente
dA
R(4
)mod
el(*
)PCSI
PFEI
CCSI
CFEI
OFEIPCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
006
1.20
4???
1.04
91.
042
1.10
6??
1.00
01.
205??
1.05
11.
049
1.10
7??
0.97
61.
206??
1.07
31.
091
1.13
10.
980
1.20
3??
1.06
31.
083
1.12
62
0.98
11.
250??
1.02
61.
066
1.13
1??
0.97
71.
250??
1.02
41.
073
1.13
1??
0.95
81.
240??
1.03
31.
117?
1.15
2?0.
961
1.23
6??
1.02
41.
108
1.14
7?
30.
982
1.22
4??
1.03
71.
065
1.12
0??
0.97
81.
225??
1.02
81.
070
1.12
1?0.
949?
1.22
6??
1.02
01.
112
1.14
1?0.
954?
1.22
0??
1.01
91.
105
1.13
8?
40.
999
1.17
1???
1.03
71.
046
1.09
0?0.
993
1.17
2???
1.03
21.
052
1.09
2?0.
956?
1.18
1???
1.03
01.
086
1.11
1?0.
964?
1.17
5???
1.03
01.
081
1.10
9?
hD
urab
leC
onsu
mpt
ion
10.
995
1.00
80.
972
0.92
30.
977
0.98
81.
007
0.96
60.
885
0.95
40.
994
1.00
60.
968
0.92
70.
974
0.98
91.
006
0.94
30.
915
0.97
02
0.98
71.
018
1.00
4?0.
982
1.02
00.
989
1.02
11.
001??
0.97
81.
012
1.01
01.
010
1.01
31.
008
1.01
01.
001
1.01
20.
989
1.00
31.
010
30.
988
1.02
10.
929
1.03
21.
035
0.98
61.
021
0.92
81.
025
1.02
51.
006
1.01
90.
964
1.04
11.
022
0.99
91.
025
0.94
81.
031
1.02
24
1.00
81.
027
0.92
21.
033
1.01
71.
004
1.03
00.
919
1.02
71.
012
1.01
51.
032
0.94
91.
032
1.00
41.
011
1.03
40.
940
1.01
81.
003
hN
on-d
urab
leC
onsu
mpt
ion
11.
022
1.18
7??
1.08
2?1.
037
1.11
2??
1.01
31.
188??
1.07
51.
044
1.11
0??
0.95
7?1.
199??
1.06
41.
086
1.12
5?0.
966
1.19
1??
1.06
71.
078
1.12
2?
20.
995
1.20
2??
1.09
21.
025
1.10
4?0.
986
1.20
4??
1.07
81.
032
1.10
3?0.
929???
1.22
1??
1.05
71.
080
1.12
5?0.
940???
1.21
3??
1.05
61.
070
1.12
0?
30.
993
1.17
7??
1.09
2?1.
019
1.09
1?0.
986
1.18
0??
1.07
71.
023
1.09
0?0.
930???
1.20
3???
1.05
31.
073
1.10
9?0.
941???
1.19
4???
1.05
31.
064
1.10
7?
41.
006
1.13
6???
1.06
7?1.
006
1.06
80.
999
1.13
9???
1.05
7?1.
012
1.06
90.
948??
1.17
2???
1.04
31.
053
1.08
8?0.
959??
1.16
1???
1.04
51.
046
1.08
6?
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
023
1.08
9??
1.04
31.
005
1.03
61.
025
1.08
4??
1.05
61.
019
1.04
51.
059?
1.06
5?1.
058
0.97
41.
023
1.05
51.
058
1.05
60.
968
1.01
42
0.98
01.
095??
0.98
11.
004
1.04
60.
987
1.10
3??
1.00
21.
026?
1.05
9?1.
045
1.08
5??
1.03
40.
991
1.03
81.
037
1.07
1??
1.02
30.
980
1.02
83
0.97
91.
124??
0.98
31.
007?
1.05
6?0.
982
1.12
6??
0.98
71.
026??
1.06
6?1.
014
1.09
6??
1.01
61.
008
1.05
11.
010
1.08
8??
1.01
50.
997
1.04
14
0.98
81.
119??
1.03
01.
012
1.05
2??
0.98
91.
115??
1.03
21.
026??
1.06
1??
1.00
41.
081??
1.05
6??
1.00
71.
044?
1.00
41.
074??
1.05
2??
0.99
81.
035
hC
onst
ruct
ion
and
Wor
ks1
1.03
41.
085?
1.08
81.
025
1.03
61.
028
1.07
9?1.
096
1.03
41.
040
1.02
31.
037
1.08
91.
030
1.02
21.
023
1.03
61.
081
1.02
71.
019
21.
008
1.07
5?1.
027
1.00
91.
032
1.00
81.
071?
1.04
31.
020
1.03
61.
013
1.03
8?1.
044
1.01
71.
021
1.01
11.
037?
1.04
01.
013
1.01
83
0.99
81.
077??
0.99
41.
003
1.03
7?1.
001
1.07
5?1.
006
1.01
11.
042?
1.00
21.
048?
1.02
01.
012
1.02
81.
001
1.04
7?1.
017
1.00
71.
025
40.
998
1.08
3??
1.01
51.
013
1.04
1??
0.99
81.
079??
1.02
01.
019?
1.04
6??
0.99
31.
050??
1.04
0?1.
024??
1.03
4?0.
994
1.05
0??
1.03
5?1.
019?
1.03
0?
hM
achi
nery
and
Equi
pmen
t1
1.01
01.
012
1.09
5??
1.05
1?1.
018
0.99
60.
982
1.09
6??
1.04
3??
1.00
61.
007
0.94
81.
070
1.02
81.
006
1.00
40.
965
1.06
81.
036
1.01
72
0.97
30.
972
1.00
71.
049???
1.01
00.
965
0.95
91.
016?
1.04
4??
1.00
61.
008
0.94
01.
024
1.02
31.
005
1.00
10.
948
1.01
21.
033?
1.01
33
0.98
70.
990
0.99
21.
030??
1.03
70.
981
0.97
90.
997
1.02
61.
032
1.01
20.
956
1.00
11.
024
1.02
61.
008
0.96
20.
999
1.02
41.
029?
40.
988
0.99
80.
996
1.03
91.
049??
0.98
70.
995
1.00
01.
040
1.05
3??
1.02
91.
010
1.01
31.
028
1.05
0??
1.02
21.
004
1.01
21.
023
1.04
7??
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
11
Tabl
e4.
A3:
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
fact
or-a
ugm
ente
dA
R(4
)mod
els
acco
rdin
gto
its
rela
tion
ship
wit
hth
ebu
sine
sscy
cle
(*)
PCSI
PFEICCSI
CFEIOFEIPCSI
PFEICCSICFEIOFEIPCSI
PFEICCSI
CFEIOFEI
PCSI
PFEICCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
989
1.05
50.
905
0.95
91.
004
0.98
11.
060
0.90
50.
966
1.00
50.
938
1.02
90.
944
1.02
81.
018
0.93
80.
987
0.92
21.
013
0.99
2
20.
949??
1.03
80.
883
0.92
10.
962
0.94
8??
1.04
20.
882
0.93
30.
965
0.88
91.
014
0.89
10.
984
0.97
80.
896?
0.99
30.
872
0.97
50.
964
30.
970
1.03
50.
927
0.94
30.
974
0.96
91.
039
0.91
90.
951
0.97
60.
900
1.03
00.
908
1.00
60.
992
0.90
81.
003
0.89
91.
001
0.98
0
40.
997
1.01
50.
951
0.94
40.
969
0.99
31.
017
0.94
80.
953
0.97
10.
917??
1.02
80.
948
1.00
60.
991
0.92
6?1.
005
0.94
71.
006
0.98
5
hD
urab
leC
onsu
mpt
ion
11.
015
0.99
00.
963
0.97
60.
988
1.00
10.
986
0.94
70.
903?
0.94
3?0.
992
0.95
50.
940
0.99
40.
967
0.99
60.
951
0.90
10.
955
0.95
9
20.
982
0.98
50.
952
0.95
00.
985
0.98
40.
990
0.94
10.
936?
0.96
91.
011
0.95
80.
959
0.99
50.
959
1.01
30.
962
0.90
7?0.
994
0.96
7
30.
993
0.98
80.
955
0.97
90.
994
0.99
10.
987
0.94
70.
964
0.97
61.
022
0.98
21.
022
0.98
50.
969
1.02
10.
981
0.98
50.
972
0.97
7
40.
996
0.98
50.
968
0.97
20.
988
0.99
00.
991
0.96
40.
963
0.97
91.
014
1.00
41.
025
0.96
30.
971
1.01
30.
999
1.00
80.
935
0.97
5
hN
on-d
urab
leC
onsu
mpt
ion
10.
975
1.06
8?0.
959
0.93
31.
004
0.96
61.
069??
0.95
60.
936
1.00
00.
901
1.06
70.
947
0.98
80.
995
0.89
91.
025
0.96
20.
977
0.97
6
20.
963??
1.05
30.
951
0.90
70.
972
0.95
5??
1.05
60.
940
0.91
60.
972
0.87
0??
1.07
7??
0.90
7?0.
975
0.98
90.
879??
1.05
10.
905
0.96
30.
978
30.
975
1.04
80.
973
0.92
00.
979
0.97
01.
052
0.96
00.
923
0.97
70.
887?
1.08
2?0.
926?
0.99
10.
992
0.89
7?1.
053
0.92
4?0.
980
0.98
3
40.
999
1.02
80.
985
0.93
00.
977
0.99
51.
031
0.97
60.
937
0.97
70.
920??
1.08
3??
0.95
70.
999
0.99
70.
930??
1.05
6??
0.95
80.
993
0.99
2
hG
ross
Fixe
dC
apit
alFo
rmat
ion
10.
997
1.02
20.
942
1.00
60.
999
0.99
41.
027
0.96
11.
035
1.02
11.
047
0.95
50.
955
0.95
20.
953
1.04
60.
909
0.95
30.
925
0.91
8
20.
979
0.99
10.
914?
0.98
60.
985
0.98
41.
011
0.93
81.
024
1.00
81.
087
0.97
70.
993
0.95
60.
952
1.08
00.
927
0.97
90.
923
0.90
7
30.
993
0.98
30.
928
0.97
10.
973
0.99
70.
996
0.92
71.
004
0.99
31.
042
0.94
30.
974
0.96
70.
943
1.04
40.
902
0.98
10.
939
0.90
6
41.
038
0.99
80.
946
0.97
30.
977
1.04
00.
998
0.94
60.
998
0.99
41.
033
0.94
00.
988
0.96
40.
953
1.04
60.
917
0.98
10.
947
0.93
2
hC
onst
ruct
ion
and
Wor
ks
11.
017
0.99
60.
934??
0.97
80.
975??
1.01
30.
992
0.94
70.
996
0.98
50.
961
0.90
00.
930
0.98
80.
942
0.96
20.
883
0.90
90.
984
0.92
9
21.
004
0.99
90.
941?
0.98
00.
980
1.00
50.
995
0.96
40.
999
0.98
90.
994
0.93
20.
972
0.99
00.
952
0.99
40.
918
0.95
80.
977
0.93
4
30.
995
0.98
70.
920?
0.97
3?0.
975
1.00
00.
987
0.93
30.
990
0.98
70.
988
0.92
60.
964
0.98
60.
944
0.98
90.
908
0.95
50.
968
0.92
3
41.
000
0.97
60.
911
0.96
2??
0.96
51.
003
0.97
50.
917
0.97
6??
0.97
60.
972
0.90
80.
960
0.98
20.
939
0.97
60.
892
0.94
50.
967
0.92
1
hM
achi
nery
and
Equi
pmen
t
10.
984
1.01
90.
959
1.01
80.
999
0.95
80.
984
0.96
31.
021
0.99
30.
949
0.86
90.
914
0.99
30.
959
0.95
40.
873
0.90
81.
006
0.96
2
20.
965
1.00
70.
924
1.04
2??
1.02
10.
947
0.99
80.
934
1.05
0?1.
024
1.00
50.
916
0.94
00.
999
0.99
01.
005
0.91
40.
914
1.01
10.
984
31.
017
1.06
20.
957
1.01
61.
049
1.00
21.
049
0.96
31.
010
1.04
31.
051
0.95
90.
967
1.00
21.
009
1.05
60.
946
0.96
00.
990
0.99
4
41.
001
0.99
80.
938
1.00
01.
020
0.99
30.
995
0.94
41.
007
1.02
91.
063
0.99
40.
964
0.97
71.
010
1.05
80.
977
0.96
40.
958
0.99
8
(?)R
elat
ive
RM
SFE2 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=
Figu
res
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Har
vey,
Leyb
ourn
e,an
dN
ewbo
ld(1
997)
test
’sp-
valu
e:(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
12
4R
obus
tnes
sre
sult
sIV
:AR
(1)n
ativ
em
odel
,1-t
erm
fact
orin
diff
eren
ces
Tabl
e1.
A4.
Adj
uste
dR
-squ
ared
rati
obe
twee
nA
R(1
)fac
tor-
augm
ente
d(1
-ter
m)a
ndna
tive
AR
(1)m
odel
(*)
and
F-te
stre
sult
s(p
-val
ue)o
fthe
cons
umer
-bas
edfa
ctor
stat
isti
cals
igni
fican
ce(*
*)Fa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sFa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEI
Fact
or:C
omm
erce
+In
dust
ryPC
0.99
51.
014
0.99
61.
002
1.00
90.
996
1.00
60.
996
0.99
81.
002
0.70
60.
170
0.59
90.
237
0.12
80.
582
0.25
60.
611
0.39
60.
254
DC
0.99
61.
026
0.99
41.
009
1.02
10.
995
1.00
80.
994
1.00
01.
005
0.42
40.
065
0.88
40.
101
0.03
10.
626
0.17
80.
677
0.26
50.
151
ND
C0.
989
1.01
60.
990
0.99
81.
007
0.99
01.
021
0.98
90.
995
1.00
50.
991
0.13
90.
778
0.32
90.
184
0.72
90.
171
0.93
40.
465
0.29
0G
FCF
0.99
91.
005
1.01
11.
002
1.00
51.
000
1.00
01.
003
1.00
61.
007
0.65
60.
304
0.18
30.
349
0.26
20.
460
0.57
00.
306
0.15
30.
177
CW
K0.
994
0.99
40.
994
0.99
80.
998
0.99
60.
994
0.99
31.
000
0.99
70.
736
0.71
80.
805
0.40
10.
413
0.48
70.
790
0.94
20.
260
0.42
8M
EQ1.
008
1.01
21.
031
1.00
51.
009
1.00
21.
009
1.01
11.
009
1.01
20.
349
0.28
50.
114
0.49
40.
357
0.71
90.
464
0.25
90.
302
0.25
9Fa
ctor
:Com
mer
ce+
Indu
stry
+Tr
ansp
orta
tion
PC0.
995
1.00
90.
997
1.00
11.
006
0.99
71.
010
0.99
50.
998
1.00
30.
724
0.19
90.
522
0.28
60.
183
0.54
00.
213
0.70
20.
361
0.20
2D
C0.
996
1.02
40.
994
1.00
61.
016
0.99
51.
008
0.99
41.
001
1.00
70.
405
0.04
60.
945
0.16
40.
054
0.65
20.
188
0.59
70.
215
0.11
7N
DC
0.98
91.
013
0.99
00.
998
1.00
60.
992
1.02
50.
989
0.99
51.
006
0.87
50.
217
0.75
30.
343
0.22
70.
560
0.12
70.
968
0.46
40.
261
GFC
F1.
001
1.00
21.
012
1.00
21.
004
0.99
81.
001
1.00
11.
007
1.00
80.
388
0.40
40.
163
0.36
60.
307
0.70
90.
507
0.36
90.
118
0.13
1C
WK
0.99
30.
994
0.99
40.
997
0.99
70.
995
0.99
30.
993
1.00
00.
997
0.96
80.
816
0.80
00.
452
0.48
90.
570
0.84
50.
907
0.20
60.
357
MEQ
1.01
21.
008
1.03
41.
005
1.00
71.
001
1.01
21.
008
1.01
01.
014
0.21
00.
408
0.08
80.
497
0.40
70.
978
0.39
10.
341
0.26
70.
207
Fact
or:C
omm
erce
+In
dust
ry+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
995
1.00
40.
996
0.99
81.
001
0.99
61.
015
0.99
51.
002
1.00
90.
695
0.28
10.
520
0.41
20.
285
0.60
50.
149
0.73
30.
236
0.11
5D
C0.
994
1.00
70.
994
1.00
11.
006
0.99
71.
022
0.99
41.
006
1.01
70.
685
0.21
40.
723
0.25
00.
145
0.36
30.
060
0.81
40.
135
0.04
4N
DC
0.98
91.
016
0.99
00.
995
1.00
40.
989
1.02
00.
989
0.99
81.
009
0.83
80.
175
0.79
30.
499
0.34
20.
858
0.09
60.
986
0.30
20.
140
GFC
F1.
000
1.00
11.
004
1.00
61.
007
0.99
81.
002
1.00
61.
003
1.00
50.
398
0.52
90.
245
0.15
10.
162
0.71
80.
424
0.26
50.
304
0.26
5C
WK
0.99
80.
994
0.99
31.
001
0.99
70.
994
0.99
50.
994
0.99
70.
997
0.37
20.
684
0.90
10.
220
0.41
30.
689
0.66
70.
806
0.43
60.
429
MEQ
1.00
21.
013
1.01
31.
008
1.01
21.
007
1.00
71.
021
1.00
71.
009
0.71
50.
390
0.20
80.
341
0.25
10.
391
0.45
90.
180
0.39
50.
351
Fact
or:C
omm
erce
+C
onst
ruct
ion
+Pe
rson
alSe
rvic
esPC
0.99
51.
009
0.99
61.
000
1.00
50.
997
1.00
50.
996
0.99
71.
001
0.68
20.
154
0.55
20.
296
0.18
50.
581
0.23
90.
668
0.42
60.
243
DC
0.99
41.
013
0.99
41.
004
1.01
10.
998
1.01
10.
994
1.00
01.
007
0.67
70.
112
0.66
80.
195
0.10
00.
303
0.11
10.
913
0.20
60.
080
ND
C0.
989
1.02
30.
989
0.99
91.
010
0.99
11.
006
0.98
90.
991
0.99
70.
939
0.11
20.
874
0.37
20.
247
0.72
20.
175
0.85
20.
565
0.29
5G
FCF
1.00
01.
002
1.00
71.
006
1.00
70.
999
1.00
01.
002
1.00
21.
003
0.46
50.
497
0.17
70.
180
0.17
20.
643
0.53
10.
427
0.32
90.
324
CW
K0.
996
0.99
40.
993
1.00
00.
997
0.99
40.
995
0.99
30.
996
0.99
70.
538
0.80
70.
973
0.22
90.
390
0.67
10.
648
0.96
10.
527
0.51
0M
EQ1.
002
1.01
41.
021
1.00
81.
013
1.00
91.
003
1.01
11.
006
1.00
60.
651
0.38
70.
119
0.35
70.
262
0.35
20.
635
0.37
30.
409
0.41
7(*
)Adj
uste
dR
-squ
ared
rati
o=A
djus
ted
R-s
quar
edw
ith
the
fact
or/A
djus
ted
R-s
quar
edw
itho
utth
efa
ctor
.Ora
nge-
shad
edce
lls=A
djus
ted
R-s
quar
edra
tio
grea
ter
than
1.G
reen
-sha
ded
cells
=Adj
uste
dR
-squ
ared
rati
ogr
eate
rth
an1.
05(5
%of
adju
stm
entg
ain)
.Sam
ple:
2001
.I-20
14.IV
(56
obse
rvat
ions
).(*
*)p-
valu
eof
the
null
hypo
thes
is:N
H:φ
1=...
=φ
4=0
(see
equa
tion
1)ag
ains
tthe
alte
rnat
ive
ofat
leas
tone
term
iseq
ual
toze
ro.G
reen
-sha
ded
cells
=p<
10%
.Ora
nge-
shad
edce
lls=p<
15%
.Sam
ple:
2001
.I-20
19.II
I(75
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
13
Tabl
e2.
A4.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(1
)mod
el,1
-ter
mfa
ctor
(*)
PCSI
PFEICCSI
CFEI
OFEI
PCSI
PFEICCSI
CFEI
OFEI
PCSI
PFEICCSI
CFEI
OFEIPCSI
PFEICCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
008
1.01
51.
025
0.99
61.
003
1.01
01.
017
1.02
60.
989
1.00
31.
001
1.01
41.
023
0.99
21.
004
1.00
61.
011
1.02
10.
989
1.00
0
21.
008
0.99
80.
994
0.98
90.
993
1.00
71.
002
0.99
40.
984
0.99
40.
998??
1.01
0?0.
998
0.98
30.
998
1.00
11.
007
0.99
90.
983
0.99
5
30.
998
1.00
31.
000
0.99
40.
999
0.99
91.
005
1.00
10.
991
0.99
90.
999
1.00
51.
001
0.99
20.
999
1.00
01.
004
1.00
10.
991
0.99
8
40.
998
1.00
21.
004
0.99
71.
000
0.99
91.
003
1.00
40.
997
1.00
00.
998
1.00
6?1.
006
0.99
81.
002
0.99
91.
005
1.00
70.
997
1.00
1
hD
urab
leC
onsu
mpt
ion
10.
999
0.99
20.
994
0.94
0??
0.96
6???
0.99
80.
995
0.99
80.
930??
0.96
5???
0.99
60.
998
1.00
30.
941??
0.97
5??
0.99
50.
997
0.99
50.
935??
0.97
0??
21.
004
0.99
51.
002
0.97
30.
987
1.00
50.
998
1.00
30.
970
0.98
81.
003
1.00
21.
001
0.97
80.
993
1.00
31.
001
1.00
20.
977
0.99
2
31.
002
1.00
01.
001
0.99
51.
001
1.00
21.
000
1.00
20.
992
0.99
91.
002
1.00
31.
001
0.99
21.
001
1.00
21.
003
1.00
20.
993
1.00
1
41.
004
1.00
01.
004
1.00
81.
006
1.00
41.
000
1.00
41.
006
1.00
51.
003
1.00
31.
003
1.00
71.
007
1.00
31.
004
1.00
61.
008
1.00
8
hN
on-d
urab
leC
onsu
mpt
ion
11.
014
1.01
91.
015
1.01
31.
019
1.01
41.
018
1.01
31.
013
1.01
81.
000
1.01
81.
015
1.01
61.
016
1.00
61.
014
1.01
61.
016
1.01
5
21.
004
1.00
40.
994
0.99
51.
004
1.00
31.
006
0.99
10.
992
1.00
30.
991?
1.01
30.
991?
0.99
61.
005
0.99
41.
009
0.99
40.
995
1.00
4
30.
998
1.00
71.
002
0.99
61.
005
0.99
81.
007
1.00
10.
994
1.00
30.
991?
1.00
70.
999
1.00
21.
004
0.99
51.
005
1.00
11.
001
1.00
4
41.
000
1.00
51.
003
0.99
51.
003
1.00
11.
005
1.00
20.
995
1.00
30.
995
1.00
71.
001
1.00
31.
005
0.99
71.
005
1.00
31.
002
1.00
5
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
001
1.00
11.
003
1.01
00.
991
1.00
3??
1.00
30.
997
1.00
60.
993
1.00
51.
007
0.99
10.
994
0.99
21.
005
1.00
60.
993
0.99
70.
992
21.
001
0.99
10.
998
0.97
80.
975?
1.00
10.
998
0.99
90.
984
0.98
41.
005
1.00
20.
993
0.96
30.
979
1.00
30.
997
0.99
20.
965
0.97
6
31.
001
0.99
80.
998
0.96
80.
983
1.00
21.
000
0.99
80.
974
0.98
81.
000
1.00
60.
995
0.96
10.
986
1.00
11.
004
0.99
60.
961
0.98
4
41.
001??
0.99
70.
995
0.97
0?0.
985?
1.00
2??
0.99
90.
995
0.97
5?0.
990?
1.00
21.
001
0.99
70.
960??
0.98
6?1.
002
1.00
00.
997
0.96
1?0.
985?
hC
onst
ruct
ion
and
Wor
ks
11.
006
1.00
61.
017???
1.01
61.
006
1.00
51.
006
1.01
7???
1.01
31.
004
1.00
31.
008???
1.01
11.
009
1.00
41.
002
1.00
7??
1.01
11.
011
1.00
4
21.
002
1.00
31.
007???
1.00
01.
000
1.00
11.
003
1.00
7???
1.00
11.
001
1.00
01.
004??
1.00
50.
999
1.00
01.
000
1.00
3??
1.00
50.
999
1.00
0
31.
001
1.00
21.
000
0.99
40.
999
1.00
11.
003
1.00
10.
996
1.00
00.
998
1.00
2?1.
001
0.99
50.
999
0.99
91.
003?
1.00
10.
995
0.99
9
41.
001
1.00
20.
999
0.99
10.
999
1.00
11.
003
1.00
00.
993
1.00
00.
999
1.00
11.
002
0.99
30.
999
0.99
91.
002
1.00
20.
992
0.99
8
hM
achi
nery
and
Equi
pmen
t
10.
998
1.00
41.
000
1.00
30.
996
1.00
11.
005
0.99
31.
000
0.99
71.
003
1.00
60.
989
0.99
40.
998
1.00
31.
007
0.99
30.
996
0.99
9
20.
999
0.99
30.
993
0.98
60.
982
1.00
00.
997
0.99
30.
989
0.98
91.
005
1.00
30.
986
0.97
70.
988
1.00
41.
000
0.98
60.
979
0.98
7
30.
998
0.99
70.
991
0.97
80.
987
1.00
00.
998
0.99
10.
980
0.99
11.
000
1.01
40.
982
0.97
30.
995
1.00
21.
013
0.98
50.
975
0.99
5
40.
998
0.99
90.
995
0.99
80.
998
1.00
01.
000
0.99
60.
999
1.00
01.
002
1.01
60.
996
0.99
51.
010
1.00
21.
016
0.99
60.
998
1.01
1
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
14
Tabl
e4.
A4:
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
1-te
rmfa
ctor
-aug
men
ted
AR
(1)m
odel
sac
cord
ing
toit
sre
lati
onsh
ipw
ith
the
busi
ness
cycl
e(*
)PCSIPFEI
CCSI
CFEIOFEIPCSIPFEI
CCSI
CFEI
OFEIPCSIPFEI
CCSI
CFEIOFEI
PCSIPFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
004
1.06
1??
1.03
21.
102
1.10
5??
1.00
51.
074??
1.03
31.
078
1.10
3??
1.04
81.
056
1.04
21.
076
1.09
3?1.
038
1.03
21.
034
1.05
81.
061?
20.
995
1.02
70.
984
1.04
31.
051
0.99
51.
038?
0.98
51.
036
1.05
3?1.
005
1.05
2??
0.99
21.
035
1.05
8?0.
997
1.03
8??
0.99
31.
032
1.04
4?
30.
990
1.03
2??
0.99
41.
053
1.04
90.
990
1.03
5?0.
997
1.05
21.
051
1.01
41.
037
1.00
41.
042
1.04
61.
009
1.02
91.
004
1.03
61.
035
40.
991
1.03
1??
0.99
61.
035
1.03
50.
991
1.03
6??
0.99
61.
036
1.03
81.
011
1.04
3??
1.00
91.
033
1.03
91.
007
1.03
3???
1.00
91.
027
1.02
9?
hD
urab
leC
onsu
mpt
ion
10.
995
0.99
40.
966??
0.97
6??
0.98
0?0.
994
0.99
70.
971??
0.96
2???
0.97
9??
0.98
60.
995
0.98
1??
0.97
30.
992
0.98
40.
990
0.96
3?0.
937??
0.97
3?
21.
001
0.99
91.
002
0.99
30.
997
1.00
21.
002
1.00
30.
986
0.99
70.
999
1.01
21.
002
0.99
51.
007
0.99
61.
008
1.00
40.
979
0.99
9
31.
000
1.00
71.
002
1.00
51.
007
1.00
01.
005
1.00
30.
998
1.00
31.
000
1.00
71.
000
0.99
81.
005
1.00
01.
006
0.99
90.
994
1.00
3
41.
002
1.00
41.
000
1.00
61.
005
1.00
21.
003
1.00
01.
003
1.00
40.
999
1.00
70.
996
1.00
51.
007
0.99
91.
009
1.00
01.
006
1.00
8
hN
on-d
urab
leC
onsu
mpt
ion
11.
003
1.08
9??
1.00
21.
037
1.07
21.
002
1.09
3???
1.00
11.
025
1.06
51.
040
1.05
71.
018
1.01
81.
033
1.03
51.
033
1.02
51.
013
1.01
9
20.
991
1.06
1??
0.97
00.
987
1.02
50.
990
1.06
6???
0.96
50.
983
1.02
50.
999
1.06
6???
0.97
10.
989
1.02
10.
997
1.04
8??
0.97
40.
988
1.01
4
30.
987
1.04
9??
0.99
00.
990
1.02
00.
987
1.05
0??
0.98
70.
987
1.01
81.
004
1.03
8?0.
997
0.99
31.
010
1.00
21.
023
1.00
10.
990
1.00
2
40.
993
1.04
9???
0.99
10.
985
1.01
60.
993
1.05
1???
0.98
80.
984
1.01
61.
009
1.04
9???
1.00
10.
992
1.01
31.
008
1.03
4??
1.00
20.
990
1.00
7
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
008
0.99
70.
991
1.01
70.
990
1.01
30.
998
0.98
31.
011
0.99
41.
020
0.98
90.
971
0.99
20.
989
1.02
40.
976
0.96
30.
994
0.98
2
21.
006
0.97
30.
968
1.00
60.
975
1.00
60.
979
0.96
71.
023??
0.99
10.
998
0.98
20.
963
0.98
40.
982
1.00
20.
977
0.95
60.
982
0.97
3
30.
997
0.99
00.
995
1.00
60.
987
0.99
90.
994
0.99
41.
020?
0.99
71.
000
1.00
10.
991
0.98
50.
989
1.00
40.
993
0.99
30.
977
0.97
8
41.
001
0.98
80.
984
0.99
90.
983?
1.00
30.
990
0.98
41.
014
0.99
21.
004
0.99
10.
988?
0.97
80.
984
1.00
60.
986
0.98
8?0.
972
0.97
7
hC
onst
ruct
ion
and
Wor
ks
11.
005
1.00
40.
999
1.01
61.
007
1.00
41.
003
0.99
91.
012
1.00
50.
998
1.00
00.
989
1.00
41.
002
0.99
70.
991
0.98
71.
009
0.99
9
21.
000
1.00
00.
996
1.00
20.
999
1.00
01.
000
0.99
61.
006
1.00
00.
996
0.99
80.
993
1.00
20.
999
0.99
60.
995
0.99
31.
002
0.99
6
30.
999
0.99
80.
995
0.99
60.
995
0.99
90.
999
0.99
61.
002
0.99
80.
996
0.99
70.
999
1.00
00.
996
0.99
70.
996
0.99
90.
998
0.99
4
41.
000
0.99
60.
991
0.99
30.
993
1.00
00.
998
0.99
20.
999
0.99
60.
996
0.99
61.
000
0.99
70.
994
0.99
70.
996
1.00
10.
993
0.99
1
hM
achi
nery
and
Equi
pmen
t
11.
003
1.00
10.
999
1.00
80.
992
1.01
01.
002
0.99
01.
004
0.99
41.
018
0.98
90.
975?
0.99
50.
991
1.02
00.
980
0.97
70.
994
0.98
6
21.
004
0.98
00.
975
1.00
50.
981
1.00
40.
983
0.97
31.
012??
0.99
11.
002
0.98
80.
965?
0.99
20.
989
1.00
40.
985
0.95
80.
992
0.98
4
30.
999
0.99
61.
001
1.01
60.
993
1.00
40.
998
0.99
71.
021
1.00
01.
011
1.01
20.
979
0.99
91.
000
1.01
40.
996
0.97
80.
992
0.98
8
40.
999
1.00
30.
980
1.01
71.
001
1.00
21.
002
0.97
91.
020
1.00
31.
006
1.00
70.
982
1.01
11.
012
1.00
80.
995
0.97
41.
012
1.00
7
(?)R
elat
ive
RM
SFE2 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=
Figu
res
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Har
vey,
Leyb
ourn
e,an
dN
ewbo
ld(1
997)
test
’sp-
valu
e:(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
15
5R
obus
tnes
sre
sult
sV
:AR
(1)n
ativ
em
odel
,1-t
erm
fact
orin
diff
eren
ces
Tabl
e1.
A5.
Adj
uste
dR
-squ
ared
rati
obe
twee
nA
R(1
)fac
tor-
augm
ente
d(2
-ter
m)a
ndna
tive
AR
(1)m
odel
(*)
and
F-te
stre
sult
s(p
-val
ue)o
fthe
cons
umer
-bas
edfa
ctor
stat
isti
cals
igni
fican
ce(*
*)Fa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sFa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEI
Fact
or:C
omm
erce
+In
dust
ryPC
0.99
41.
018
0.99
61.
007
1.01
40.
995
1.00
51.
006
1.00
41.
007
0.86
00.
222
0.78
10.
282
0.16
40.
800
0.48
10.
276
0.37
30.
285
DC
0.99
01.
029
0.99
01.
012
1.02
50.
989
1.00
31.
000
1.00
01.
005
0.74
20.
065
0.74
30.
195
0.07
40.
902
0.43
70.
358
0.38
40.
282
ND
C0.
979
1.02
30.
980
1.00
01.
012
0.98
51.
011
1.00
40.
996
1.00
40.
960
0.22
80.
941
0.35
20.
219
0.56
90.
394
0.33
00.
413
0.36
8G
FCF
0.99
61.
013
1.01
01.
014
1.01
90.
995
0.99
91.
023
1.02
71.
026
0.75
90.
242
0.37
00.
144
0.07
60.
715
0.55
60.
074
0.04
80.
057
CW
K0.
987
0.98
80.
987
0.99
20.
991
0.99
10.
988
0.98
60.
994
0.99
00.
927
0.91
20.
968
0.68
10.
708
0.71
00.
916
0.99
20.
504
0.70
4M
EQ1.
005
1.03
01.
037
1.04
51.
051
0.99
31.
017
1.07
71.
073
1.07
60.
510
0.14
90.
199
0.06
00.
022
0.90
30.
246
0.01
10.
025
0.01
6Fa
ctor
:Com
mer
ce+
Indu
stry
+Tr
ansp
orta
tion
PC0.
995
1.01
50.
998
1.00
61.
013
0.99
51.
007
1.00
51.
004
1.00
70.
812
0.16
90.
595
0.31
50.
186
0.80
40.
491
0.33
60.
354
0.26
4D
C0.
990
1.02
80.
992
1.00
81.
020
0.98
91.
003
1.00
01.
001
1.00
60.
734
0.04
80.
666
0.27
90.
126
0.91
90.
461
0.31
50.
324
0.23
6N
DC
0.98
01.
019
0.98
30.
997
1.00
80.
988
1.01
51.
004
0.99
81.
005
0.93
80.
248
0.85
50.
438
0.29
90.
490
0.32
40.
330
0.34
30.
300
GFC
F0.
997
1.00
81.
011
1.01
31.
016
0.99
31.
000
1.02
51.
029
1.02
90.
656
0.32
10.
319
0.16
40.
105
0.89
00.
520
0.05
90.
037
0.04
2C
WK
0.98
70.
987
0.98
70.
991
0.99
00.
991
0.98
80.
987
0.99
50.
991
0.99
00.
961
0.96
60.
737
0.77
70.
738
0.92
70.
984
0.42
40.
629
MEQ
1.00
51.
024
1.04
01.
042
1.04
70.
991
1.02
01.
084
1.08
01.
083
0.41
10.
208
0.14
60.
075
0.03
40.
996
0.22
70.
006
0.02
10.
011
Fact
or:C
omm
erce
+In
dust
ry+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
994
1.00
21.
006
1.00
41.
006
0.99
51.
021
0.99
61.
007
1.01
50.
813
0.54
70.
267
0.37
00.
314
0.79
80.
155
0.75
90.
299
0.16
1D
C0.
988
1.00
20.
999
1.00
11.
006
0.99
11.
025
0.99
21.
009
1.02
00.
933
0.49
20.
376
0.35
70.
254
0.66
50.
056
0.63
30.
255
0.12
6N
DC
0.98
61.
006
1.00
20.
998
1.00
20.
979
1.03
50.
982
0.99
71.
013
0.53
80.
407
0.34
30.
369
0.39
30.
950
0.15
10.
892
0.39
70.
205
GFC
F0.
995
0.99
91.
026
1.02
81.
027
0.99
51.
012
1.00
71.
014
1.01
90.
660
0.56
90.
055
0.03
60.
044
0.84
90.
282
0.39
80.
154
0.09
5C
WK
0.99
30.
989
0.98
60.
995
0.99
10.
988
0.98
80.
987
0.99
10.
991
0.57
70.
869
0.99
40.
433
0.66
90.
921
0.89
30.
967
0.73
00.
732
MEQ
0.99
31.
018
1.08
31.
079
1.08
11.
002
1.02
81.
035
1.04
41.
050
0.91
80.
233
0.00
80.
016
0.00
90.
562
0.19
10.
188
0.07
80.
037
Fact
or:C
omm
erce
+C
onst
ruct
ion
+Pe
rson
alSe
rvic
esPC
0.99
41.
007
1.00
51.
005
1.00
90.
995
1.02
20.
996
1.00
51.
012
0.82
50.
363
0.26
50.
327
0.25
80.
797
0.07
10.
810
0.31
00.
150
DC
0.98
81.
008
0.99
81.
005
1.01
10.
993
1.02
70.
991
1.00
31.
014
0.93
30.
335
0.41
40.
298
0.20
50.
578
0.03
90.
668
0.31
60.
172
ND
C0.
985
1.01
30.
996
1.00
11.
008
0.98
01.
035
0.98
40.
992
1.00
50.
643
0.28
60.
457
0.32
90.
314
0.92
30.
095
0.84
80.
523
0.29
0G
FCF
0.99
51.
001
1.02
41.
028
1.02
70.
996
1.01
01.
004
1.00
91.
013
0.72
10.
506
0.06
40.
043
0.04
40.
776
0.31
80.
567
0.21
70.
144
CW
K0.
991
0.98
80.
987
0.99
50.
991
0.98
80.
991
0.98
60.
990
0.99
20.
738
0.91
00.
990
0.44
90.
640
0.90
40.
744
0.99
80.
798
0.76
3M
EQ0.
993
1.02
41.
083
1.08
11.
083
1.00
61.
016
1.02
01.
027
1.03
20.
881
0.20
90.
010
0.01
40.
008
0.45
40.
332
0.35
30.
156
0.09
7(*
)Adj
uste
dR
-squ
ared
rati
o=A
djus
ted
R-s
quar
edw
ith
the
fact
or/A
djus
ted
R-s
quar
edw
itho
utth
efa
ctor
.Ora
nge-
shad
edce
lls=A
djus
ted
R-s
quar
edra
tio
grea
ter
than
1.G
reen
-sha
ded
cells
=Adj
uste
dR
-squ
ared
rati
ogr
eate
rth
an1.
05(5
%of
adju
stm
entg
ain)
.Sam
ple:
2001
.I-20
14.IV
(56
obse
rvat
ions
).(*
*)p-
valu
eof
the
null
hypo
thes
is:N
H:φ
1=...
=φ
4=0
(see
equa
tion
1)ag
ains
tthe
alte
rnat
ive
ofat
leas
tone
term
iseq
ual
toze
ro.G
reen
-sha
ded
cells
=p<
10%
.Ora
nge-
shad
edce
lls=p<
15%
.Sam
ple:
2001
.I-20
19.II
I(75
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
16
Tabl
e2.
A5.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(1
)mod
el,2
-ter
mfa
ctor
(*)
PCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
997
1.00
10.
912
0.96
70.
983
0.99
41.
014
0.91
30.
945
0.98
10.
992
1.02
50.
949
0.93
70.
983
0.99
21.
017
0.95
30.
934
0.97
7
20.
994
1.05
0??
0.94
21.
029
1.03
5??
0.99
31.
051???
0.93
51.
025
1.03
3??
0.97
8?1.
046?
0.93
81.
045
1.03
9??
0.98
1?1.
045??
0.94
41.
038
1.03
5??
30.
986
1.06
4???
0.97
51.
019
1.03
1?0.
986
1.05
7???
0.96
91.
017
1.02
8?0.
974?
1.04
1??
0.95
41.
042
1.03
4?0.
976?
1.04
5??
0.95
81.
036
1.03
3
40.
986
1.03
6???
0.95
00.
985
0.99
90.
984
1.03
3???
0.94
50.
982
0.99
90.
974
1.02
3??
0.94
10.
992
1.00
20.
977
1.02
5??
0.94
50.
988
1.00
1
hD
urab
leC
onsu
mpt
ion
11.
001
0.97
10.
954
0.96
30.
972
0.99
60.
975
0.95
10.
943
0.96
30.
991
0.99
50.
960
0.96
40.
979
0.98
90.
992
0.95
30.
955
0.97
5
21.
002
0.98
51.
009
1.02
21.
008
1.00
10.
990
1.00
11.
026
1.00
80.
999
0.99
70.
993
1.05
0?1.
013
0.99
90.
996
0.99
31.
043?
1.01
2
31.
002
1.00
00.
993
1.01
11.
003
1.00
00.
999
0.98
71.
006
0.99
91.
000
0.99
90.
980
1.01
31.
001
1.00
11.
000
0.98
01.
011
1.00
1
41.
005
0.99
80.
985
0.99
60.
992
1.00
30.
998
0.98
10.
993
0.99
11.
003
0.99
80.
972
0.98
90.
992
1.00
30.
999
0.97
30.
991
0.99
3
hN
on-d
urab
leC
onsu
mpt
ion
10.
985
1.03
00.
961
0.96
70.
998
0.98
21.
029
0.95
90.
944
0.99
10.
970
1.03
00.
959
0.93
70.
985
0.97
01.
023
0.96
70.
935
0.97
9
20.
970
1.05
3?0.
957
0.96
00.
998
0.96
91.
052
0.94
60.
950
0.99
70.
934??
1.06
20.
919??
0.97
01.
012
0.93
6??
1.05
90.
927??
0.96
31.
007
30.
975
1.04
6??
0.97
70.
946
0.99
40.
974
1.04
0??
0.97
10.
944
0.99
40.
943?
1.03
9?0.
942
0.96
31.
002
0.94
8?1.
041?
0.94
90.
958
1.00
2
40.
982
1.02
7???
0.96
30.
923
0.97
70.
980
1.02
3???
0.95
6?0.
922
0.97
80.
959
1.02
3???
0.93
6?0.
926
0.98
10.
963
1.02
4???
0.94
2?0.
923
0.97
9
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
005
1.00
61.
026
1.00
50.
985
1.00
81.
013
1.02
41.
005
0.99
21.
001
1.01
51.
013
0.99
00.
988
1.00
41.
008
1.01
20.
993
0.98
5
21.
000
0.99
71.
015
0.97
60.
972?
0.99
91.
006
1.01
80.
981
0.98
21.
001
0.99
91.
000
0.94
70.
963?
1.00
10.
989
0.99
50.
949
0.95
5??
31.
006
1.00
30.
990
0.97
80.
989
1.00
41.
008
0.99
30.
975
0.99
41.
005
1.01
0??
0.98
10.
942
0.97
91.
005
1.00
50.
980
0.94
30.
972
41.
002
1.00
50.
992
0.97
60.
990
1.00
01.
008
0.99
20.
975
0.99
50.
999
1.00
7?0.
987?
0.94
3??
0.98
2?0.
999
1.00
40.
985?
0.94
4??
0.97
7?
hC
onst
ruct
ion
and
Wor
ks
11.
014?
1.02
01.
042??
1.01
91.
011
1.01
11.
019
1.04
1??
1.02
01.
011
1.00
71.
015???
1.02
41.
013
1.00
61.
007
1.01
4??
1.02
51.
013
1.00
5
21.
007
1.01
61.
015??
0.99
41.
001
1.00
51.
016
1.01
7??
1.00
11.
004
1.00
31.
009?
1.01
30.
994
1.00
11.
003
1.00
9?1.
012
0.99
30.
998
31.
006
1.01
30.
998
0.98
11.
002
1.00
61.
015
1.00
00.
987
1.00
51.
001
1.00
9??
1.00
60.
983
1.00
11.
002
1.00
9??
1.00
50.
980
0.99
8
41.
005??
1.01
00.
997
0.97
21.
001
1.00
4?1.
012?
0.99
90.
978
1.00
40.
999
1.00
8?1.
010?
0.97
30.
998
1.00
01.
008??
1.00
8?0.
969
0.99
6
hM
achi
nery
and
Equi
pmen
t
11.
004
1.01
11.
011
1.00
10.
995
1.00
61.
014
1.01
00.
998
0.99
80.
996
1.01
61.
003
0.99
51.
001
1.00
21.
015
1.00
20.
996
1.00
1
20.
996
0.99
71.
009
0.99
00.
982??
0.99
61.
000
1.01
50.
987
0.99
00.
996
0.99
90.
993
0.96
80.
982?
0.99
70.
995
0.98
90.
971
0.97
7?
30.
995
1.00
60.
968
0.99
10.
996
0.99
41.
005
0.97
70.
984
0.99
90.
991
1.01
6?0.
957
0.97
10.
997
0.99
51.
016
0.95
80.
974
0.99
2
40.
997
1.00
4??
0.97
70.
999
0.99
80.
995?
1.00
20.
978
0.99
60.
999
0.99
61.
013
0.98
00.
990
1.00
70.
998
1.01
80.
981
0.99
31.
007
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
17
Tabl
e4.
A5:
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
2-te
rmfa
ctor
-aug
men
ted
AR
(1)m
odel
sac
cord
ing
toit
sre
lati
onsh
ipw
ith
the
busi
ness
cycl
e(*
)PCSIPFEI
CCSICFEIOFEIPCSIPFEI
CCSICFEIOFEIPCSIPFEICCSI
CFEIOFEI
PCSIPFEICCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
001
1.02
10.
957
1.17
21.
118?
0.99
71.
053
0.95
31.
110
1.11
1?1.
042
1.05
21.
039
1.06
11.
089
1.01
80.
999
1.06
01.
021
1.02
5
21.
005
1.01
90.
977
0.99
31.
008
1.00
71.
026
0.96
60.
980
1.00
30.
982
1.01
20.
958
1.02
21.
015
0.97
61.
000
0.95
81.
007
1.00
2
31.
005
1.02
50.
998
0.98
91.
002
1.00
51.
016
0.98
90.
982
0.99
60.
983
0.99
10.
953
1.02
81.
003
0.98
50.
998
0.95
31.
019
1.00
3
40.
995
1.01
30.
983
1.01
41.
004
0.99
11.
013
0.97
31.
012
1.00
70.
982
0.99
40.
967
1.02
91.
006
0.98
30.
993
0.97
11.
013
0.99
7
hD
urab
leC
onsu
mpt
ion
11.
010
0.97
80.
987
0.98
40.
985
1.00
10.
982
0.98
20.
947
0.96
80.
981
1.01
40.
994
0.97
10.
991
0.97
30.
999
0.98
10.
928?
0.96
5
21.
008
1.00
11.
029
0.94
50.
983
1.00
41.
008
1.01
70.
944
0.98
20.
996
1.02
10.
994
0.98
90.
990
0.99
11.
012
0.99
30.
968
0.98
1
31.
005
1.02
01.
027
0.99
21.
009
1.00
21.
019
1.01
80.
979
1.00
11.
001
1.01
40.
997
0.99
31.
000
1.00
11.
012
0.98
90.
985
0.99
9
41.
003
1.01
91.
021
1.02
61.
022
1.00
01.
019
1.01
51.
019
1.02
01.
000
1.01
60.
991
1.01
11.
017
1.00
01.
019
0.98
81.
011
1.02
1
hN
on-d
urab
leC
onsu
mpt
ion
10.
984
1.06
8???
0.95
91.
068
1.06
60.
980
1.07
3???
0.95
81.
014
1.04
71.
008
1.03
90.
968
0.97
00.
998
0.98
81.
008
0.99
80.
950
0.96
2
20.
992
1.04
5???
0.97
10.
984
0.98
90.
995
1.04
4???
0.95
50.
964
0.98
60.
941
1.05
8??
0.90
5???
1.00
11.
009
0.94
01.
052??
0.90
7??
0.98
20.
997
30.
993
1.03
6??
0.99
50.
975
0.98
60.
993
1.02
8?0.
985
0.96
90.
985
0.95
11.
023
0.93
7??
1.00
40.
994
0.95
41.
025
0.94
5?0.
991
0.98
9
40.
994
1.03
6?0.
991
0.99
20.
993
0.99
31.
030?
0.98
00.
989
0.99
50.
973
1.03
00.
951?
0.99
20.
994
0.97
41.
026
0.95
50.
974
0.98
2
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
012
0.99
40.
988
1.01
60.
989
1.01
51.
006
0.98
71.
017
1.00
21.
011
0.98
70.
970
0.99
20.
989
1.01
20.
955
0.95
2?0.
992
0.96
8
21.
003
0.97
90.
985
1.02
50.
988
1.00
10.
992
0.99
01.
039??
1.00
80.
993
0.96
60.
965
0.97
00.
969
0.99
40.
939
0.94
50.
958
0.93
1?
30.
997
1.00
01.
028
1.05
1??
1.01
30.
996
1.00
91.
032
1.05
0?1.
023?
1.00
51.
006
1.00
80.
963
0.98
41.
008
0.98
30.
999
0.94
20.
947
41.
002
0.99
01.
003
1.03
4?0.
998
1.00
10.
995
1.00
51.
036
1.00
80.
998
0.98
7?0.
992
0.95
8??
0.97
8?1.
001
0.97
40.
982?
0.94
1??
0.95
1?
hC
onst
ruct
ion
and
Wor
ks
11.
002
1.00
81.
014
1.01
31.
001
0.99
81.
008
1.01
11.
019
1.00
40.
990
0.98
70.
979
1.00
60.
989
0.99
00.
974
0.98
11.
009
0.98
1
20.
996
1.00
60.
997
0.99
70.
991
0.99
41.
008
0.99
71.
012
0.99
60.
995
0.98
50.
993
1.00
10.
987
0.99
60.
978
0.99
30.
997
0.97
3
30.
997
0.99
90.
988
0.98
40.
985
0.99
61.
003
0.98
90.
999
0.99
00.
996
0.98
91.
006
0.99
30.
981
0.99
90.
985
1.00
80.
982
0.96
7
41.
000
0.98
80.
978
0.97
8?0.
978
0.99
90.
993
0.97
8.9
920.
984?
0.99
30.
986
1.00
70.
986
0.97
30.
995
0.98
21.
008
0.97
0?0.
959
hM
achi
nery
and
Equi
pmen
t
11.
008
1.00
10.
994
1.00
10.
988
1.01
11.
008
0.99
60.
994
0.99
21.
012
0.99
20.
980
0.99
30.
994
1.01
20.
971
0.96
70.
989
0.98
2
21.
006
0.97
90.
984
1.01
40.
985
1.00
60.
983
0.99
61.
015
0.99
90.
994
0.97
00.
964
0.97
80.
982
0.99
50.
960
0.94
4?0.
974
0.96
0
31.
004
0.99
81.
043
1.03
7?1.
006
1.00
40.
999
1.06
01.
019
1.01
11.
003
1.00
31.
018
0.98
50.
998
1.01
10.
984
1.01
30.
972
0.96
8
41.
007
0.98
90.
994
1.02
30.
995
1.00
50.
985
0.99
51.
015
0.99
60.
998
0.98
30.
999
1.00
71.
004
1.00
50.
978
0.99
41.
005
0.99
4
(?)R
elat
ive
RM
SFE2 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=
Figu
res
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Har
vey,
Leyb
ourn
e,an
dN
ewbo
ld(1
997)
test
’sp-
valu
e:(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
18
6R
obus
tnes
sre
sult
sV
I:A
R(1
)nat
ive
mod
el,3
-ter
mfa
ctor
indi
ffer
ence
s
Tabl
e1.
A6.
Adj
uste
dR
-squ
ared
rati
obe
twee
nA
R(1
)fac
tor-
augm
ente
d(3
-ter
m)a
ndna
tive
AR
(1)m
odel
(*)
and
F-te
stre
sult
s(p
-val
ue)o
fthe
cons
umer
-bas
edfa
ctor
stat
isti
cals
igni
fican
ce(*
*)Fa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sFa
ctor
:Mor
ese
nsit
ive
sect
ors
Fact
or:L
ess
sens
itiv
ese
ctor
sPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEI
Fact
or:C
omm
erce
+In
dust
ryPC
0.98
51.
010
0.98
90.
999
1.00
70.
985
0.99
61.
012
0.99
50.
999
0.94
30.
345
0.73
70.
400
0.23
30.
938
0.64
30.
098
0.53
10.
442
DC
0.98
11.
024
0.98
21.
010
1.02
40.
981
0.99
70.
997
0.99
30.
999
0.89
30.
139
0.80
50.
138
0.05
10.
948
0.55
10.
394
0.41
20.
301
ND
C0.
968
1.00
90.
962
0.98
40.
998
0.96
70.
994
0.99
80.
980
0.98
90.
927
0.41
50.
975
0.54
70.
351
0.80
20.
609
0.37
90.
563
0.49
0G
FCF
0.99
01.
008
1.01
01.
007
1.01
10.
989
0.99
31.
020
1.02
01.
018
0.91
60.
428
0.29
30.
316
0.19
10.
893
0.77
90.
108
0.13
30.
148
CW
K0.
978
0.99
50.
979
0.99
20.
995
0.98
30.
986
0.97
80.
991
0.99
00.
984
0.55
90.
980
0.63
50.
564
0.87
00.
857
0.99
20.
496
0.56
1M
EQ0.
997
1.02
11.
049
1.03
61.
041
0.98
51.
009
1.08
01.
064
1.06
60.
735
0.33
50.
102
0.18
50.
080
0.97
20.
458
0.02
80.
085
0.05
8Fa
ctor
:Com
mer
ce+
Indu
stry
+Tr
ansp
orta
tion
PC0.
984
1.00
70.
996
0.99
81.
005
0.98
50.
998
1.00
80.
996
0.99
90.
931
0.23
70.
442
0.45
70.
277
0.93
60.
690
0.18
00.
501
0.40
3D
C0.
981
1.02
40.
986
1.00
41.
018
0.98
20.
997
0.99
50.
995
1.00
10.
881
0.09
00.
622
0.22
80.
096
0.95
40.
610
0.41
90.
341
0.25
0N
DC
0.96
61.
004
0.96
60.
982
0.99
40.
971
0.99
80.
997
0.98
30.
992
0.94
30.
473
0.90
80.
649
0.48
20.
724
0.53
20.
393
0.44
60.
376
GFC
F0.
991
1.00
21.
013
1.00
61.
009
0.98
70.
995
1.02
11.
022
1.02
10.
837
0.54
70.
236
0.34
40.
247
0.98
00.
706
0.11
00.
110
0.11
3C
WK
0.97
70.
992
0.97
80.
989
0.99
10.
982
0.98
60.
978
0.99
40.
993
0.99
80.
656
0.99
30.
733
0.65
00.
893
0.84
70.
975
0.37
70.
446
MEQ
0.99
71.
016
1.05
21.
032
1.03
70.
983
1.01
11.
084
1.07
01.
073
0.63
50.
400
0.08
80.
206
0.11
30.
998
0.44
30.
019
0.07
70.
045
Fact
or:C
omm
erce
+In
dust
ry+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
985
0.99
61.
011
0.99
50.
997
0.98
81.
013
0.99
20.
998
1.00
70.
876
0.58
70.
076
0.51
00.
449
0.83
10.
300
0.69
90.
441
0.26
0D
C0.
979
0.99
90.
995
0.99
51.
002
0.98
41.
018
0.98
61.
005
1.01
60.
987
0.53
90.
395
0.37
30.
238
0.78
80.
122
0.69
10.
201
0.11
8N
DC
0.96
80.
989
0.99
40.
983
0.98
80.
976
1.02
60.
967
0.98
10.
999
0.72
90.
597
0.37
70.
496
0.53
70.
813
0.27
30.
941
0.60
00.
314
GFC
F0.
990
0.99
31.
023
1.02
11.
019
0.99
01.
006
1.01
01.
007
1.01
20.
837
0.78
40.
101
0.10
90.
124
0.94
50.
482
0.25
40.
311
0.21
9C
WK
0.98
40.
989
0.97
80.
991
0.99
10.
978
0.98
80.
979
0.99
20.
994
0.78
20.
786
0.98
80.
439
0.51
60.
985
0.71
50.
978
0.65
00.
606
MEQ
0.98
61.
009
1.08
31.
069
1.07
20.
995
1.01
81.
051
1.03
41.
041
0.96
30.
443
0.02
50.
059
0.03
80.
793
0.39
20.
084
0.21
80.
118
Fact
or:C
omm
erce
+C
onst
ruct
ion
+Pe
rson
alSe
rvic
esPC
0.98
51.
000
1.01
10.
997
1.00
10.
989
1.01
40.
990
0.99
51.
004
0.92
50.
399
0.06
10.
447
0.34
00.
830
0.19
90.
807
0.51
80.
304
DC
0.97
91.
006
0.99
41.
000
1.00
90.
987
1.02
00.
985
0.99
61.
007
0.98
60.
429
0.44
20.
269
0.15
80.
698
0.09
50.
690
0.35
70.
253
ND
C0.
966
0.99
70.
988
0.98
60.
996
0.98
21.
028
0.96
90.
975
0.99
00.
837
0.45
90.
447
0.45
70.
436
0.74
20.
246
0.93
40.
735
0.46
4G
FCF
0.98
90.
995
1.02
31.
021
1.02
00.
991
1.00
51.
004
1.00
31.
007
0.88
60.
749
0.07
90.
122
0.12
10.
915
0.44
90.
442
0.41
20.
293
CW
K0.
982
0.99
00.
978
0.99
30.
993
0.97
90.
988
0.97
80.
988
0.98
90.
894
0.75
70.
985
0.43
30.
474
0.98
00.
702
0.98
40.
768
0.74
7M
EQ0.
986
1.01
51.
088
1.07
11.
073
0.99
91.
008
1.03
21.
018
1.02
30.
960
0.39
70.
016
0.05
30.
031
0.70
30.
527
0.19
60.
364
0.23
7(*
)Adj
uste
dR
-squ
ared
rati
o=A
djus
ted
R-s
quar
edw
ith
the
fact
or/A
djus
ted
R-s
quar
edw
itho
utth
efa
ctor
.Ora
nge-
shad
edce
lls=A
djus
ted
R-s
quar
edra
tio
grea
ter
than
1.G
reen
-sha
ded
cells
=Adj
uste
dR
-squ
ared
rati
ogr
eate
rth
an1.
05(5
%of
adju
stm
entg
ain)
.Sam
ple:
2001
.I-20
14.IV
(56
obse
rvat
ions
).(*
*)p-
valu
eof
the
null
hypo
thes
is:N
H:φ
1=...
=φ
4=0
(see
equa
tion
1)ag
ains
tthe
alte
rnat
ive
ofat
leas
tone
term
iseq
ual
toze
ro.G
reen
-sha
ded
cells
=p<
10%
.Ora
nge-
shad
edce
lls=p<
15%
.Sam
ple:
2001
.I-20
19.II
I(75
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
19
Tabl
e2.
A6.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(1
)mod
el,3
-ter
mfa
ctor
(*)
PCSI
PFEICCSI
CFEIOFEI
PCSI
PFEICCSI
CFEIOFEI
PCSI
PFEICCSI
CFEI
OFEI
PCSI
PFEICCSI
CFEI
OFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
993
1.01
50.
957
0.97
80.
992
0.99
41.
029?
0.96
60.
954
0.98
60.
993
1.03
00.
982
0.95
40.
992
0.99
01.
019
0.98
10.
952
0.98
3
20.
965
1.10
8??
0.96
11.
030
1.05
7?0.
969
1.10
4??
0.95
51.
026
1.05
4?0.
960?
1.09
0??
0.94
21.
073??
1.06
9?0.
958
1.09
3??
0.93
81.
062?
1.06
5?
30.
969
1.15
7???
0.97
61.
032
1.07
0??
0.96
81.
143???
0.96
71.
037
1.06
7??
0.95
0?1.
125???
0.93
91.
081??
1.07
9??
0.95
2?1.
134???
0.93
71.
070?
1.07
7??
40.
977
1.09
7???
0.94
20.
987
1.01
40.
974
1.09
2???
0.93
70.
987
1.01
50.
960
1.08
7???
0.91
31.
014
1.02
20.
961
1.09
0???
0.91
41.
006
1.01
9
hD
urab
leC
onsu
mpt
ion
11.
005
0.97
20.
971
0.94
20.
954
1.00
20.
978
0.96
70.
928
0.95
00.
996
0.99
40.
959
0.95
70.
973
0.99
30.
990
0.95
00.
949
0.96
8
21.
001
0.98
51.
006
1.00
70.
998
1.00
10.
984
0.99
51.
010
0.99
41.
003
0.98
60.
978
1.04
41.
000
1.00
00.
986
0.97
31.
038
0.99
8
31.
008
0.99
20.
980
1.01
10.
992
1.00
50.
986
0.97
11.
008
0.98
61.
006
0.98
10.
951
1.02
00.
984
1.00
40.
983
0.94
91.
016
0.98
5
41.
009
0.98
50.
964
0.97
60.
966
1.00
70.
984
0.95
80.
971
0.96
31.
009
0.98
10.
924
0.96
70.
957
1.00
80.
981
0.92
70.
971
0.96
0
hN
on-d
urab
leC
onsu
mpt
ion
10.
980
1.04
20.
987
0.97
31.
002
0.98
01.
040
0.99
10.
950
0.99
30.
963??
1.03
60.
985
0.95
10.
988
0.96
4??
1.03
00.
989
0.94
90.
984
20.
951??
1.09
2?0.
995
0.94
41.
006
0.95
0???
1.09
1?0.
982
0.93
51.
001
0.91
3???
1.10
1?0.
942
0.97
71.
018
0.91
6???
1.10
1?0.
941
0.96
61.
015
30.
961??
1.09
8??
1.02
60.
932
1.00
50.
958??
1.09
5??
1.01
40.
929
1.00
00.
917??
1.10
5???
0.96
10.
968
1.01
00.
925??
1.10
6???
0.95
90.
959
1.00
8
40.
977
1.06
6???
1.00
10.
901
0.97
90.
975
1.06
5???
0.99
00.
898
0.97
70.
944?
1.08
1???
0.94
90.
921
0.98
00.
950?
1.07
9???
0.94
70.
913
0.97
9
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
019
0.99
11.
051
1.01
60.
987
1.02
2?1.
010
1.04
51.
018
0.99
91.
026
1.00
61.
008
1.00
00.
988
1.02
8?0.
997
1.00
81.
006
0.98
7
21.
008
1.00
31.
041
0.98
91.
002
1.00
91.
019
1.04
50.
999
1.01
41.
022
1.01
31.
011
0.98
41.
004
1.02
01.
004
1.01
00.
985
0.99
7
31.
011??
1.03
41.
014
0.99
21.
044??
1.01
01.
040
1.01
70.
994
1.04
6??
1.02
11.
050
1.02
11.
010
1.05
8???
1.02
01.
050
1.02
41.
003
1.05
3???
41.
006??
1.02
51.
016
0.98
61.
028???
1.00
4?1.
029
1.01
60.
990
1.03
1???
1.00
6??
1.03
51.
024?
0.99
91.
040??
1.00
6?1.
034
1.02
40.
993
1.03
5??
hC
onst
ruct
ion
and
Wor
ks
11.
018
1.03
31.
044
1.02
21.
019
1.01
61.
031
1.04
91.
026
1.02
01.
013
1.02
0??
1.03
41.
022
1.01
41.
013
1.02
0??
1.03
31.
022
1.01
2
21.
009
1.02
71.
009
0.98
51.
007
1.00
91.
027
1.01
50.
992
1.01
01.
007
1.01
71.
020
0.98
61.
005
1.00
71.
018
1.01
80.
984
1.00
1
31.
007
1.02
60.
992
0.94
51.
005
1.00
71.
027
0.99
40.
952
1.00
61.
000
1.01
81.
016
0.95
10.
999
1.00
11.
019
1.01
20.
944
0.99
5
41.
003
1.02
11.
002
0.93
21.
004
1.00
31.
022
1.00
10.
941
1.00
70.
995
1.01
81.
026?
0.94
20.
999
0.99
61.
018
1.02
1?0.
932
0.99
4
hM
achi
nery
and
Equi
pmen
t
11.
017??
0.99
21.
045?
1.01
80.
993
1.01
9???
1.00
61.
038
1.01
71.
001
1.01
90.
998
1.00
31.
012
0.99
61.
022?
0.99
51.
004
1.01
60.
998
21.
006
0.98
31.
032?
1.00
41.
004
1.00
60.
997
1.03
5?1.
009
1.01
61.
017
0.99
80.
996
1.00
61.
011
1.01
50.
990
0.99
51.
009
1.00
8
31.
005
1.04
11.
023
1.03
51.
072?
1.00
51.
044
1.03
21.
036?
1.07
6?1.
018?
1.06
21.
017
1.07
9??
1.10
4??
1.01
9?1.
067
1.02
4?1.
075??
1.10
2??
41.
002
1.01
20.
994
1.01
31.
021
1.00
11.
009
0.99
81.
014
1.02
21.
011
1.00
91.
007
1.03
6?1.
039
1.01
21.
016
1.01
01.
035?
1.03
9
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
20
Tabl
e4.
A6:
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
3-te
rmfa
ctor
-aug
men
ted
AR
(2)m
odel
sac
cord
ing
toit
sre
lati
onsh
ipw
ith
the
busi
ness
cycl
e(*
)PCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
994
1.03
10.
989
1.16
81.
116?
0.99
41.
069??
0.99
41.
113
1.10
7?1.
040
1.05
51.
058
1.06
11.
086
1.01
31.
000
1.06
81.
024
1.02
4
20.
973
1.05
90.
976
0.98
11.
017
0.98
31.
064??
0.97
00.
973
1.01
30.
963
1.03
30.
937
1.03
91.
030
0.95
31.
018
0.92
61.
020
1.01
7
30.
982
1.08
6??
0.97
00.
997
1.02
60.
983
1.07
2?0.
963
1.00
21.
022
0.95
21.
035
0.91
41.
064
1.03
00.
957
1.04
10.
900
1.05
01.
027
40.
983
1.05
4?0.
955
0.99
71.
009
0.98
01.
052
0.95
10.
997
1.01
10.
966
1.03
80.
921
1.03
51.
019
0.96
51.
033
0.91
21.
016
1.00
8
hD
urab
leC
onsu
mpt
ion
11.
013
0.98
41.
007
0.97
30.
979
1.00
70.
994
1.00
20.
941
0.96
90.
988
1.01
70.
982
0.98
41.
012
0.97
51.
003
0.96
10.
961
0.98
9
21.
012
1.01
61.
049
0.94
00.
990
1.01
11.
017
1.03
10.
935
0.98
41.
009
1.01
60.
985
1.00
30.
994
1.00
11.
010
0.96
90.
993
0.98
5
31.
011
1.03
5?1.
048
0.98
41.
011
1.00
71.
028
1.03
20.
973
1.00
01.
011
1.00
70.
982
0.99
50.
993
1.00
71.
009
0.96
70.
983
0.99
3
41.
008
1.03
1?1.
057
1.02
11.
029
1.00
51.
031?
1.04
91.
010
1.02
41.
014
1.01
90.
970
1.00
81.
013
1.01
01.
021
0.96
41.
012
1.01
7
hN
on-d
urab
leC
onsu
mpt
ion
10.
973
1.06
8???
0.97
51.
056
1.05
90.
972
1.07
2???
0.98
01.
002
1.03
70.
993
1.03
10.
998
0.96
30.
988
0.97
31.
005
1.02
90.
944
0.95
7
20.
965??
1.06
7???
0.96
2???
0.94
80.
986
0.96
7??
1.06
8???
0.94
8???
0.92
9?0.
978
0.91
6?1.
079??
0.89
7??
0.99
41.
005
0.91
3?1.
074??
0.88
7?0.
972
0.99
5
30.
972?
1.06
8??
0.96
80.
930
0.98
70.
970?
1.06
5??
0.95
70.
920
0.97
90.
921
1.07
2???
0.89
2?0.
984
0.98
90.
926
1.06
6???
0.87
8?0.
968
0.98
2
40.
983
1.06
0???
0.96
80.
936
0.98
90.
981
1.06
0???
0.95
60.
927
0.98
60.
954
1.08
3???
0.91
40.
965
0.98
80.
958
1.07
1???
0.90
00.
943
0.97
7
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
002
0.98
51.
013
1.00
90.
987
1.00
21.
019
1.00
91.
012
1.00
91.
023
0.98
20.
938?
0.98
40.
981
1.01
80.
938
0.92
5???
0.98
80.
958
20.
995
0.98
31.
006
0.99
90.
993
0.99
41.
009
1.01
61.
017
1.01
51.
014
0.99
00.
957?
0.99
40.
993
1.01
10.
959
0.94
8?0.
986
0.96
3
30.
999
0.99
60.
989
1.00
61.
006
0.99
81.
003
0.99
01.
007
1.00
91.
020
1.01
70.
996
1.04
0?1.
029?
1.02
31.
019
1.00
11.
018
1.01
6
41.
004
0.98
40.
983
0.99
90.
991
1.00
20.
989
0.98
21.
003
0.99
61.
004
0.99
60.
995
1.02
31.
011
1.00
70.
996
0.99
41.
005
1.00
0
hC
onst
ruct
ion
and
Wor
ks
11.
002
1.01
61.
010
1.00
41.
000
1.00
01.
016
1.01
61.
013
1.00
20.
996
0.98
60.
988
1.00
60.
987
0.99
60.
972
0.99
01.
010
0.97
4
21.
000
1.01
20.
981
1.00
30.
997
1.00
01.
015
0.98
61.
018
1.00
11.
000
0.99
00.
998
1.01
10.
989
1.00
40.
987
1.00
01.
004
0.97
3
31.
001
1.00
70.
967
0.99
60.
996
1.00
01.
010
0.96
41.
011
1.00
10.
996
0.98
91.
007
1.01
7?0.
984
1.00
00.
987
1.00
40.
996
0.96
6
41.
001
0.99
20.
961??
0.98
40.
988
1.00
10.
996
0.95
4??
1.00
20.
995
0.98
90.
986
1.00
71.
015??
0.97
80.
990
0.98
31.
002
0.98
50.
958
hM
achi
nery
and
Equi
pmen
t
11.
000
0.98
01.
015
0.99
20.
981
1.00
01.
009
1.00
80.
989
0.99
71.
018
0.96
40.
936??
0.98
70.
979
1.01
20.
934
0.92
5???
0.99
00.
966
20.
988
0.97
51.
020
0.98
00.
984
0.98
60.
997
1.02
90.
988
1.00
31.
009
0.99
00.
958
0.98
60.
990
1.00
30.
961
0.94
9?0.
987
0.97
2
31.
003
0.94
90.
993
0.96
40.
964
1.00
30.
951
1.00
80.
962
0.96
81.
021
0.97
20.
982
1.04
3?1.
015
1.02
70.
979
0.99
61.
036
1.01
4
41.
008
0.98
20.
978
0.99
00.
982
1.00
50.
978
0.98
20.
989
0.98
21.
015
0.96
01.
003
1.04
21.
011
1.02
00.
960
1.00
21.
041
1.01
1
(?)R
elat
ive
RM
SFE2 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=
Figu
res
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Har
vey,
Leyb
ourn
e,an
dN
ewbo
ld(1
997)
test
’sp-
valu
e:(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
21
7R
obus
tnes
sre
sult
sV
II:A
R(1
)nat
ive
mod
el,1
-ter
mfa
ctor
inle
vels
Tabl
e1.
A7.
Adj
uste
dR
-squ
ared
rati
obe
twee
nA
R(1
)fac
tor-
augm
ente
d(1
-ter
m,i
nle
vels
)and
nati
veA
R(1
)mod
el(*
)an
dF-
test
resu
lts
(p-v
alue
)oft
heco
nsum
er-b
ased
fact
orst
atis
tica
lsig
nific
ance
(**)
Fact
or:M
ore
sens
itiv
ese
ctor
sFa
ctor
:Les
sse
nsit
ive
sect
ors
Fact
or:M
ore
sens
itiv
ese
ctor
sFa
ctor
:Les
sse
nsit
ive
sect
ors
PCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEI
Fact
or:C
omm
erce
+In
dust
ryPC
0.99
81.
005
1.01
30.
996
0.99
60.
996
0.99
81.
008
0.99
80.
996
0.52
00.
248
0.07
40.
807
0.82
60.
814
0.49
80.
135
0.49
20.
727
DC
0.99
51.
014
1.00
30.
995
0.99
70.
995
1.00
11.
000
0.99
70.
995
0.84
60.
132
0.38
80.
989
0.61
70.
873
0.43
90.
314
0.54
00.
847
ND
C0.
998
1.00
71.
008
0.99
20.
993
0.99
40.
999
0.99
70.
993
0.99
20.
514
0.30
80.
150
0.94
20.
777
0.73
50.
450
0.51
50.
743
0.94
4G
FCF
0.99
61.
007
0.99
41.
000
1.00
40.
999
0.99
60.
994
1.00
21.
002
0.54
50.
186
0.91
20.
319
0.20
50.
273
0.54
50.
841
0.19
30.
211
CW
K0.
994
1.00
00.
995
1.00
61.
007
0.99
60.
994
0.99
61.
001
0.99
90.
911
0.44
30.
742
0.28
20.
263
0.48
70.
936
0.60
70.
355
0.46
2M
EQ0.
995
1.00
30.
990
0.99
10.
995
0.99
50.
995
0.99
00.
995
0.99
70.
400
0.26
20.
958
0.68
30.
491
0.46
80.
466
0.84
90.
435
0.38
3Fa
ctor
:Com
mer
ce+
Indu
stry
+Tr
ansp
orta
tion
PC0.
998
1.00
21.
012
0.99
60.
996
0.99
61.
000
1.00
70.
997
0.99
60.
492
0.29
40.
075
0.74
20.
963
0.86
60.
432
0.14
40.
492
0.77
7D
C0.
995
1.01
21.
003
0.99
50.
996
0.99
51.
001
0.99
90.
997
0.99
50.
792
0.12
60.
361
0.86
40.
747
0.89
30.
436
0.34
40.
569
0.89
2N
DC
0.99
91.
003
1.00
70.
992
0.99
30.
993
1.00
10.
997
0.99
30.
992
0.44
90.
363
0.18
40.
891
0.87
90.
839
0.40
30.
536
0.75
60.
993
GFC
F0.
998
1.00
20.
994
1.00
01.
002
0.99
70.
998
0.99
41.
003
1.00
40.
293
0.29
10.
916
0.34
10.
259
0.44
30.
444
0.80
80.
159
0.16
0C
WK
0.99
40.
997
0.99
51.
004
1.00
40.
995
0.99
40.
995
1.00
10.
999
0.83
80.
531
0.69
00.
320
0.31
70.
589
0.99
90.
640
0.32
30.
419
MEQ
1.00
00.
998
0.99
00.
992
0.99
40.
992
0.99
80.
990
0.99
60.
999
0.24
90.
414
0.91
20.
677
0.54
80.
648
0.37
00.
784
0.40
00.
321
Fact
or:C
omm
erce
+In
dust
ry+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
996
0.99
81.
009
0.99
80.
996
0.99
71.
004
1.00
90.
996
0.99
60.
783
0.50
90.
118
0.47
10.
696
0.57
70.
286
0.11
20.
790
0.85
6D
C0.
995
1.00
21.
000
0.99
70.
995
0.99
61.
008
1.00
20.
995
0.99
60.
957
0.41
20.
322
0.58
00.
915
0.75
90.
190
0.36
70.
872
0.75
6N
DC
0.99
40.
997
1.00
00.
993
0.99
20.
997
1.00
81.
001
0.99
20.
993
0.71
60.
481
0.43
00.
748
0.92
50.
576
0.29
40.
311
0.91
00.
790
GFC
F0.
998
0.99
70.
994
1.00
21.
003
0.99
61.
003
0.99
41.
000
1.00
30.
336
0.46
90.
963
0.18
60.
188
0.46
70.
292
0.84
30.
301
0.22
9C
WK
0.99
60.
994
0.99
51.
001
0.99
90.
994
0.99
70.
996
1.00
51.
005
0.53
90.
960
0.70
90.
339
0.44
10.
953
0.51
60.
625
0.30
50.
304
MEQ
0.99
30.
998
0.99
00.
995
0.99
80.
995
0.99
80.
990
0.99
20.
995
0.53
00.
376
0.98
30.
455
0.36
50.
397
0.41
00.
817
0.60
70.
491
Fact
or:C
omm
erce
+C
onst
ruct
ion
+Pe
rson
alSe
rvic
esPC
0.99
61.
000
1.00
90.
996
0.99
60.
997
1.00
01.
011
0.99
70.
996
0.71
00.
359
0.12
20.
656
0.92
10.
691
0.42
80.
085
0.50
10.
826
DC
0.99
51.
005
0.99
90.
996
0.99
50.
995
1.00
31.
005
0.99
70.
995
0.83
30.
331
0.35
50.
792
0.88
00.
918
0.29
00.
298
0.52
20.
929
ND
C0.
996
1.00
11.
000
0.99
20.
993
0.99
41.
000
1.00
30.
994
0.99
20.
611
0.32
30.
429
0.89
80.
888
0.74
20.
431
0.22
20.
627
0.90
1G
FCF
0.99
80.
999
0.99
41.
003
1.00
50.
996
0.99
90.
995
0.99
81.
000
0.35
00.
408
0.86
90.
174
0.15
90.
461
0.38
40.
655
0.41
40.
343
CW
K0.
995
0.99
40.
995
1.00
21.
000
0.99
40.
998
0.99
51.
003
1.00
40.
637
0.95
20.
648
0.31
00.
394
0.93
80.
444
0.73
50.
331
0.30
9M
EQ0.
994
1.00
00.
990
0.99
60.
999
0.99
60.
992
0.99
40.
990
0.99
10.
488
0.34
60.
815
0.42
20.
325
0.40
70.
621
0.48
60.
791
0.70
2(*
)Adj
uste
dR
-squ
ared
rati
o=A
djus
ted
R-s
quar
edw
ith
the
fact
or/A
djus
ted
R-s
quar
edw
itho
utth
efa
ctor
.Ora
nge-
shad
edce
lls=A
djus
ted
R-s
quar
edra
tio
grea
ter
than
1.G
reen
-sha
ded
cells
=Adj
uste
dR
-squ
ared
rati
ogr
eate
rth
an1.
05(5
%of
adju
stm
entg
ain)
.Sam
ple:
2001
.I-20
14.IV
(56
obse
rvat
ions
).(*
*)p-
valu
eof
the
null
hypo
thes
is:N
H:φ
1=...
=φ
4=0
(see
equa
tion
1)ag
ains
tthe
alte
rnat
ive
ofat
leas
tone
term
iseq
ual
toze
ro.G
reen
-sha
ded
cells
=p<
10%
.Ora
nge-
shad
edce
lls=p<
15%
.Sam
ple:
2001
.I-20
19.II
I(75
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
22
Tabl
e2.
A7.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(1
)mod
el(1
-ter
m,i
nle
vels
)(*)
PCSI
PFEICCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
010
1.05
2???
1.01
0?1.
010
1.04
1??
1.00
31.
050???
1.00
7??
1.01
21.
041??
0.98
61.
043?
1.02
8??
1.02
61.
041?
0.98
91.
044?
1.02
6??
1.02
31.
041?
20.
979
1.09
6??
0.95
41.
009
1.05
3???
0.97
41.
096??
0.95
1?1.
014
1.05
4???
0.95
71.
088??
1.00
1??
1.03
8?1.
065??
0.95
91.
089??
0.99
8??
1.03
1?1.
061??
30.
974
1.13
4??
0.99
01.
048
1.06
5??
0.96
91.
132??
0.98
61.
052
1.06
6??
0.94
21.
114??
1.03
01.
086?
1.08
0?0.
945
1.11
5??
1.02
21.
075?
1.07
6?
40.
984
1.13
3??
0.99
31.
071
1.05
2??
0.97
61.
129??
0.99
11.
077
1.05
5?0.
942
1.11
7??
1.04
01.
107?
1.07
2?0.
947
1.12
0??
1.03
61.
095?
1.06
8?
hD
urab
leC
onsu
mpt
ion
11.
003
0.99
31.
006
1.00
51.
003
1.00
20.
993
1.00
61.
005?
1.00
2?1.
001
0.99
61.
019
1.00
81.
002
1.00
10.
996
1.01
61.
006
1.00
3
21.
002
0.98
60.
989
1.01
10.
999
1.00
10.
986
0.98
91.
012
0.99
81.
001
0.99
31.
016
1.01
50.
999
1.00
10.
994
1.01
11.
013
1.00
1
31.
001
0.98
50.
983
1.01
00.
996
1.00
00.
985
0.98
11.
012
0.99
41.
000
0.99
21.
005
1.01
50.
995
1.00
10.
993
1.00
01.
013
0.99
9
41.
004
0.98
30.
975
1.00
90.
995
1.00
20.
983
0.97
31.
011
0.99
11.
000
0.99
20.
992
1.01
40.
992
1.00
10.
993
0.99
01.
011
0.99
7
hN
on-d
urab
leC
onsu
mpt
ion
11.
002
1.07
4???
0.99
51.
001
1.02
5??
0.99
61.
072???
0.98
81.
005
1.02
7??
0.96
81.
070??
1.00
11.
026
1.03
60.
973
1.07
0?1.
001
1.02
01.
035
20.
969
1.12
5??
0.96
30.
970
1.02
1??
0.96
31.
125??
0.95
10.
974
1.02
4?0.
923
1.12
5???
0.96
21.
014
1.04
4?0.
929
1.12
2??
0.96
41.
002
1.03
9?
30.
969???
1.14
1??
0.98
80.
979
1.02
40.
962???
1.13
8??
0.97
70.
982
1.02
70.
914?
1.13
3???
0.98
21.
027
1.04
80.
922??
1.13
2???
0.98
11.
014
1.04
4
40.
979
1.14
3???
0.98
20.
991
1.02
40.
971
1.14
1???
0.97
20.
995
1.02
90.
917??
1.14
0???
0.97
91.
037
1.05
00.
927??
1.13
9???
0.98
11.
025
1.04
7
hG
ross
Fixe
dC
apit
alFo
rmat
ion
10.
999
1.00
31.
004
1.00
00.
994
0.99
91.
004
1.00
30.
999
0.99
61.
001
1.00
21.
000
0.99
50.
994
1.00
11.
002
0.99
90.
994
0.99
2
21.
002
1.00
61.
000
0.99
30.
990
1.00
21.
008
1.00
00.
991
0.99
61.
002
1.00
80.
997
0.98
00.
992??
1.00
21.
007
0.99
70.
978
0.98
7??
31.
000
1.00
50.
980
0.98
00.
989
1.00
11.
008
0.98
00.
979
0.99
71.
003
1.00
90.
978
0.95
90.
990?
1.00
21.
007
0.97
90.
955
0.98
3?
41.
000
1.00
4?0.
990
0.97
90.
993
1.00
11.
003?
0.98
90.
978
0.99
91.
003
1.00
20.
992
0.96
10.
994?
1.00
11.
001
0.99
10.
956
0.98
8??
hC
onst
ruct
ion
and
Wor
ks
11.
000?
1.00
91.
005
0.99
51.
003
1.00
0??
1.00
91.
005
0.99
51.
004
0.99
91.
008?
1.00
50.
993
0.99
90.
999
1.00
7?1.
004
0.99
20.
997
21.
003
1.00
71.
007
0.98
11.
004
1.00
21.
008
1.00
80.
984
1.00
50.
998
1.00
61.
012
0.98
21.
000
0.99
91.
006
1.01
10.
979
0.99
8
31.
003
1.00
71.
007
0.97
21.
007
1.00
21.
008
1.00
90.
979
1.00
90.
998
1.00
51.
019
0.97
61.
004
0.99
81.
006
1.01
70.
971
1.00
1
41.
003
1.01
01.
015
0.97
01.
010
1.00
21.
010
1.01
60.
976
1.01
30.
996
1.00
6?1.
029??
0.97
51.
007
0.99
61.
007
1.02
6??
0.96
91.
004
hM
achi
nery
and
Equi
pmen
t
11.
004
1.00
51.
011?
1.01
11.
002
1.00
41.
000
1.01
1?1.
010
0.99
81.
010
0.99
41.
009?
1.01
61.
000
1.00
8?1.
001
1.00
81.
017
1.00
4
21.
004
1.00
61.
016
1.01
80.
999
1.00
41.
004
1.01
61.
016
0.99
31.
011
1.00
41.
010
1.02
30.
998
1.00
91.
011
1.01
11.
025
1.00
4
31.
005
1.03
5?0.
986
1.03
70.
996?
1.00
61.
033??
0.98
61.
030
0.99
0?1.
021
1.02
61.
000
1.02
60.
996?
1.01
61.
034
0.99
91.
031
1.00
1
41.
007
0.98
60.
996
1.02
2??
0.99
41.
008
0.97
80.
996
1.01
5??
0.98
61.
028
0.99
31.
001
1.02
7?0.
997
1.02
21.
007
1.00
01.
028?
1.00
4
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
23
Tabl
e4.
A7:
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
fact
or-a
ugm
ente
d(1
-ter
m,i
nle
vels
)AR
(1)m
odel
sac
cord
ing
toit
sre
lati
onsh
ipw
ith
the
busi
ness
cycl
e(*
)PCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
11.
008
1.05
1??
0.99
61.
003
1.03
31.
001
1.04
8??
0.99
21.
005
1.03
30.
978??
1.03
6?1.
018
1.02
11.
030?
0.98
1?1.
033??
1.01
81.
018
1.02
7
20.
982
1.07
9???
0.94
0?0.
992
1.02
70.
976?
1.08
0???
0.93
50.
997
1.02
80.
950??
1.06
0??
1.00
01.
028
1.03
90.
951??
1.05
7??
0.99
91.
019
1.03
0
30.
975
1.09
9???
0.95
81.
018
1.02
60.
971
1.09
9???
0.95
21.
022
1.02
80.
934??
1.06
7??
1.00
61.
065??
1.03
90.
937??
1.06
2??
0.99
61.
049?
1.03
2
40.
984
1.09
7???
0.95
41.
038
1.01
30.
976
1.09
4???
0.95
01.
045
1.01
70.
928??
1.06
7??
1.01
31.
080??
1.03
10.
934??
1.06
5??
1.01
21.
062??
1.02
6
hD
urab
leC
onsu
mpt
ion
11.
001
1.00
11.
002
0.99
91.
006??
0.99
91.
000
1.00
20.
998
1.00
5?0.
998
1.00
51.
024??
1.00
41.
004
0.99
81.
007
1.02
3?1.
001
1.00
5
21.
000
0.99
60.
993
0.99
81.
005
0.99
80.
996
0.99
20.
999
1.00
41.
001
1.00
51.
041??
1.00
41.
005
1.00
11.
008
1.04
0??
1.00
01.
010
30.
998
0.99
41.
001
0.99
61.
009
0.99
70.
993
0.99
80.
997
1.00
61.
002
1.00
71.
041?
1.00
11.
007
1.00
21.
010
1.03
80.
991
1.01
4?
41.
000
0.99
51.
005
0.99
21.
012
0.99
70.
993
1.00
20.
993
1.00
71.
000
1.00
91.
039
0.99
61.
009
1.00
11.
012
1.04
40.
982
1.01
9
hN
on-d
urab
leC
onsu
mpt
ion
11.
000
1.06
0???
0.98
50.
990
1.01
20.
994
1.06
0???
0.97
80.
993
1.01
40.
962?
1.05
6???
0.99
11.
017
1.02
20.
966?
1.05
4???
0.99
31.
010
1.02
0
20.
970??
1.09
4???
0.95
3??
0.95
50.
995
0.96
5??
1.09
4???
0.93
9??
0.95
80.
997
0.92
1??
1.09
6???
0.94
7?1.
006
1.02
00.
925??
1.08
8???
0.95
2?0.
993
1.01
3
30.
970??
1.10
5???
0.96
80.
960
0.99
30.
963??
1.10
4???
0.95
50.
961
0.99
60.
912??
1.10
1???
0.95
71.
015
1.01
90.
920??
1.09
4???
0.95
71.
001
1.01
5
40.
978
1.10
9???
0.96
00.
971
0.99
20.
971?
1.10
7???
0.94
90.
973
0.99
50.
912???
1.10
8???
0.95
31.
020
1.01
80.
920???
1.10
2???
0.95
71.
007
1.01
5
hG
ross
Fixe
dC
apit
alFo
rmat
ion
10.
997
1.01
21.
005
1.00
40.
999
0.99
71.
014
1.00
31.
001
1.00
01.
002
1.00
90.
995
0.99
70.
997
1.00
41.
012
0.99
1?0.
996
0.99
4
20.
997
1.02
20.
995
1.00
00.
990
0.99
51.
023
0.99
30.
997
0.99
50.
997
1.01
70.
991
0.98
10.
988
0.99
91.
022
0.99
20.
977
0.98
0
30.
994
1.03
10.
978
0.99
10.
987
0.99
41.
031
0.97
50.
990
0.99
31.
002
1.02
10.
975
0.96
10.
984
1.00
11.
024
0.97
60.
952
0.97
2
40.
994
1.03
60.
980
0.98
40.
995
0.99
41.
032
0.97
70.
982
0.99
61.
001
1.02
20.
984
0.96
40.
992
0.99
91.
028
0.98
3?0.
955
0.98
5
hC
onst
ruct
ion
and
Wor
ks
11.
000
1.00
50.
993
1.00
11.
000
1.00
01.
006
0.99
21.
001
1.00
10.
996
1.00
40.
994
0.99
90.
992
0.99
81.
003
0.99
20.
996
0.98
4
21.
001
1.01
30.
988
0.99
30.
999
1.00
01.
014
0.98
60.
999
1.00
10.
995
1.01
00.
997
0.99
60.
991
0.99
61.
009
0.99
50.
989?
0.98
3
31.
001
1.02
0?0.
979
0.98
60.
999
1.00
01.
019?
0.97
80.
995
1.00
20.
995
1.01
40.
999
0.99
3?0.
991
0.99
51.
013
0.99
60.
981??
0.98
2
41.
002
1.02
10.
973
0.98
00.
998
1.00
21.
020
0.97
10.
989?
1.00
20.
993
1.01
10.
996
0.99
2?0.
990
0.99
11.
011
0.99
00.
977??
0.98
2
hM
achi
nery
and
Equi
pmen
t
10.
989
1.02
91.
019
1.00
61.
043??
0.98
81.
024
1.02
01.
001
1.03
1??
1.01
01.
016
1.00
51.
015
1.03
1?1.
008
1.03
31.
002
1.02
2?1.
044?
20.
980
1.00
81.
039?
1.01
61.
063???
0.97
81.
010
1.04
0?1.
009
1.05
0???
1.01
01.
014
1.02
11.
024
1.04
6???
1.00
61.
024
1.02
31.
038
1.06
6???
30.
971
1.00
40.
989
1.04
31.
099??
0.96
81.
004
0.98
91.
027
1.07
6??
1.01
40.
984
1.00
71.
030
1.06
2??
1.00
91.
002
1.00
11.
053
1.08
0?
40.
954
1.01
71.
029
1.03
31.
094
0.95
01.
008
1.02
81.
014
1.07
01.
027
1.04
01.
016
1.03
91.
072
1.01
61.
075
1.01
41.
052
1.09
4
(?)R
elat
ive
RM
SFE2 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=
Figu
res
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Har
vey,
Leyb
ourn
e,an
dN
ewbo
ld(1
997)
test
’sp-
valu
e:(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
24
8R
obus
tnes
sre
sult
sV
III:
AR
(1)n
ativ
em
odel
,2-t
erm
fact
orin
leve
ls
Tabl
e1.
A8.
Adj
uste
dR
-squ
ared
rati
obe
twee
nA
R(1
)fac
tor-
augm
ente
d(2
-ter
m,i
nle
vels
)and
nati
veA
R(1
)mod
el(*
)an
dF-
test
resu
lts
(p-v
alue
)oft
heco
nsum
er-b
ased
fact
orst
atis
tica
lsig
nific
ance
(**)
Fact
or:M
ore
sens
itiv
ese
ctor
sFa
ctor
:Les
sse
nsit
ive
sect
ors
Fact
or:M
ore
sens
itiv
ese
ctor
sFa
ctor
:Les
sse
nsit
ive
sect
ors
PCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEI
Fact
or:C
omm
erce
+In
dust
ryPC
1.00
61.
011
1.01
71.
008
1.01
40.
998
1.00
51.
012
1.00
81.
012
0.34
00.
309
0.09
40.
149
0.08
90.
598
0.35
70.
209
0.16
00.
126
DC
1.00
01.
023
1.01
21.
017
1.02
70.
995
1.00
41.
006
1.01
41.
017
0.37
00.
166
0.16
20.
066
0.02
20.
700
0.29
30.
142
0.12
10.
083
ND
C0.
997
1.00
91.
010
0.99
91.
008
0.99
41.
024
0.99
31.
003
1.01
70.
443
0.26
40.
179
0.36
40.
211
0.49
20.
162
0.53
50.
396
0.22
5G
FCF
0.99
31.
005
1.01
30.
998
1.00
20.
996
0.99
41.
003
1.00
21.
002
0.81
40.
408
0.18
30.
533
0.39
40.
591
0.82
10.
291
0.29
00.
337
CW
K0.
988
0.99
60.
988
0.99
81.
000
0.99
00.
987
0.99
20.
995
0.99
20.
934
0.63
40.
935
0.56
80.
540
0.74
50.
942
0.60
60.
541
0.73
2M
EQ0.
998
1.00
41.
042
0.99
40.
998
0.99
50.
998
1.01
00.
999
1.00
20.
623
0.52
80.
103
0.78
70.
643
0.75
30.
754
0.27
20.
565
0.51
0Fa
ctor
:Com
mer
ce+
Indu
stry
+Tr
ansp
orta
tion
PC1.
004
1.00
61.
016
1.00
81.
011
0.99
91.
010
1.01
31.
008
1.01
40.
370
0.40
20.
114
0.16
30.
122
0.57
40.
258
0.18
10.
150
0.10
6D
C1.
001
1.02
01.
010
1.01
61.
023
0.99
41.
005
1.00
71.
014
1.01
80.
377
0.12
30.
188
0.08
20.
031
0.70
00.
300
0.11
20.
121
0.07
9N
DC
0.99
61.
007
1.00
61.
001
1.00
90.
998
1.02
90.
994
1.00
21.
017
0.46
50.
370
0.24
50.
338
0.21
80.
389
0.10
90.
512
0.41
80.
233
GFC
F0.
996
0.99
91.
015
0.99
81.
000
0.99
40.
996
1.00
11.
003
1.00
40.
574
0.59
30.
163
0.57
10.
485
0.73
80.
723
0.33
40.
223
0.24
3C
WK
0.98
70.
993
0.98
80.
997
0.99
70.
989
0.98
70.
992
0.99
60.
992
0.95
00.
724
0.90
00.
615
0.60
30.
834
0.94
80.
575
0.46
50.
667
MEQ
1.00
30.
998
1.04
40.
994
0.99
70.
994
1.00
21.
006
1.00
01.
005
0.43
30.
708
0.08
20.
792
0.70
50.
771
0.64
10.
325
0.50
70.
417
Fact
or:C
omm
erce
+In
dust
ry+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
996
1.00
21.
012
1.00
91.
012
1.00
71.
013
1.01
51.
007
1.01
40.
674
0.46
60.
236
0.16
10.
148
0.30
80.
218
0.10
70.
143
0.07
3D
C0.
992
1.00
21.
005
1.01
51.
017
1.00
41.
021
1.01
31.
015
1.02
50.
812
0.39
90.
195
0.11
80.
090
0.24
80.
141
0.10
60.
080
0.02
6N
DC
0.99
21.
019
0.99
51.
003
1.01
60.
998
1.01
41.
004
1.00
01.
011
0.56
70.
211
0.51
30.
428
0.27
10.
418
0.16
30.
289
0.32
30.
156
GFC
F0.
995
0.99
51.
002
1.00
11.
002
0.99
41.
000
1.01
20.
999
1.00
10.
656
0.76
50.
333
0.28
10.
301
0.67
80.
557
0.16
10.
498
0.42
7C
WK
0.99
10.
989
0.99
10.
996
0.99
20.
990
0.99
10.
989
0.99
70.
998
0.67
50.
837
0.71
90.
484
0.71
20.
834
0.78
90.
812
0.59
70.
593
MEQ
0.99
31.
002
1.00
90.
998
1.00
20.
998
0.99
81.
035
0.99
60.
998
0.83
10.
673
0.28
90.
607
0.49
10.
630
0.72
30.
115
0.69
20.
639
Fact
or:C
omm
erce
+C
onst
ruct
ion
+Pe
rson
alSe
rvic
esPC
0.99
81.
008
1.01
21.
009
1.01
41.
004
1.00
31.
019
1.00
51.
008
0.58
00.
268
0.21
80.
157
0.12
40.
416
0.36
70.
082
0.19
10.
127
DC
0.99
41.
009
1.00
41.
015
1.02
01.
003
1.01
01.
018
1.01
31.
018
0.71
10.
213
0.19
60.
134
0.08
70.
258
0.21
80.
072
0.07
50.
027
ND
C0.
993
1.02
50.
996
1.00
51.
019
0.99
61.
000
1.00
40.
996
1.00
10.
555
0.17
10.
466
0.37
90.
237
0.47
00.
270
0.26
70.
467
0.27
6G
FCF
0.99
50.
997
1.00
41.
002
1.00
40.
995
0.99
71.
011
0.99
70.
998
0.68
80.
710
0.25
60.
293
0.28
10.
680
0.66
10.
200
0.59
20.
572
CW
K0.
989
0.98
90.
990
0.99
60.
993
0.99
00.
992
0.98
90.
997
0.99
70.
829
0.87
00.
760
0.49
20.
678
0.82
80.
694
0.81
20.
614
0.57
8M
EQ0.
994
1.00
41.
016
0.99
91.
003
0.99
90.
993
1.03
30.
996
0.99
60.
813
0.65
00.
190
0.60
10.
480
0.60
70.
890
0.13
20.
682
0.69
1(*
)Adj
uste
dR
-squ
ared
rati
o=A
djus
ted
R-s
quar
edw
ith
the
fact
or/A
djus
ted
R-s
quar
edw
itho
utth
efa
ctor
.Ora
nge-
shad
edce
lls=A
djus
ted
R-s
quar
edra
tio
grea
ter
than
1.G
reen
-sha
ded
cells
=Adj
uste
dR
-squ
ared
rati
ogr
eate
rth
an1.
05(5
%of
adju
stm
entg
ain)
.Sam
ple:
2001
.I-20
14.IV
(56
obse
rvat
ions
).(*
*)p-
valu
eof
the
null
hypo
thes
is:N
H:φ
1=...
=φ
4=0
(see
equa
tion
1)ag
ains
tthe
alte
rnat
ive
ofat
leas
tone
term
iseq
ual
toze
ro.G
reen
-sha
ded
cells
=p<
10%
.Ora
nge-
shad
edce
lls=p<
15%
.Sam
ple:
2001
.I-20
19.II
I(75
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
25
Tabl
e2.
A8.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(1
)mod
el(2
-ter
m,i
nle
vels
)(*)
PCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEI
OFEI
PCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
899
0.87
30.
916
0.88
10.
850
0.91
60.
919
0.92
30.
843
0.83
70.
997
0.89
01.
011
0.86
20.
841
0.96
40.
861
0.99
70.
857
0.82
9
20.
932
1.08
41.
009
0.96
11.
024
0.93
81.
082?
1.01
90.
966
1.02
50.
953
1.02
51.
080?
1.06
11.
046
0.93
71.
027
1.06
1?1.
045
1.04
3
30.
913
1.17
0??
0.98
31.
027
1.06
00.
902
1.16
1??
0.99
91.
047
1.06
60.
901
1.12
21.
102
1.13
0?1.
096
0.89
21.
128
1.08
41.
109?
1.09
1
40.
909
1.17
6??
0.93
01.
098?
1.04
60.
902
1.17
4??
0.94
91.
113??
1.05
60.
903
1.13
4?1.
077?
1.18
4???
1.08
90.
890
1.13
6?1.
059
1.15
9??
1.07
8
hD
urab
leC
onsu
mpt
ion
10.
992
0.95
60.
912
0.90
20.
896???
0.99
00.
964
0.91
60.
883
0.88
9???
0.99
80.
989
0.94
90.
907
0.92
6??
0.99
20.
985
0.93
70.
901
0.92
1??
21.
006
0.98
10.
983
0.97
10.
941??
1.01
00.
985
0.98
60.
972
0.94
3??
1.02
3?0.
989
1.03
70.
986
0.95
7?1.
020?
0.98
81.
029
0.98
40.
957
31.
008
0.97
90.
969
1.03
60.
960
1.00
80.
977
0.97
61.
036
0.95
7?1.
024?
0.98
91.
035
1.03
40.
967?
1.02
3?0.
989
1.03
01.
032
0.96
8
41.
011
0.97
40.
954
1.06
90.
955?
1.01
10.
976
0.96
01.
076
0.95
6?1.
031
0.98
61.
021
1.07
40.
967?
1.02
90.
984
1.01
81.
069
0.96
6?
hN
on-d
urab
leC
onsu
mpt
ion
10.
908
0.95
00.
999
0.96
40.
952
0.91
60.
951
1.00
00.
940
0.92
90.
982
0.85
3?1.
018
0.95
20.
895
0.95
80.
838?
1.01
40.
945
0.88
9
20.
915?
1.13
3???
1.01
10.
958
1.01
80.
911??
1.12
4???
1.01
00.
970
1.02
00.
903??
1.05
2???
1.02
81.
065?
1.04
30.
896??
1.06
3???
1.01
61.
037
1.03
9
30.
908??
1.20
9???
0.96
50.
964
1.03
50.
898???
1.20
1???
0.96
80.
975
1.04
10.
874???
1.14
6???
0.99
91.
065
1.07
30.
871???
1.15
1???
0.99
21.
035
1.06
5
40.
894???
1.22
2???
0.92
00.
989
1.04
00.
887???
1.21
6???
0.92
30.
996
1.04
50.
857???
1.17
3???
0.97
01.
084?
1.08
30.
855???
1.17
8???
0.96
21.
054
1.07
4
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
041
0.99
21.
013
1.02
70.
997
1.03
7?0.
998
1.00
61.
030
1.00
31.
038
1.01
31.
002
1.00
41.
001
1.03
61.
013
1.00
91.
010
1.00
4
21.
013
0.97
10.
981
0.99
40.
989
1.01
60.
982
0.98
41.
001
1.00
11.
039
0.98
90.
979
0.97
00.
987
1.03
00.
982
0.97
70.
973
0.98
3
31.
010
1.00
31.
010
0.98
71.
016
1.01
21.
002
1.01
30.
992
1.02
01.
017
1.00
71.
012
0.98
21.
013
1.01
51.
005
1.01
00.
980
1.01
1
41.
009
0.98
91.
010
0.97
81.
004
1.01
10.
989
1.01
10.
984
1.00
91.
021
0.99
31.
013
0.97
01.
001
1.01
50.
993
1.01
00.
968
1.00
0
hC
onst
ruct
ion
and
Wor
ks
11.
029
1.03
21.
040
1.04
5?1.
044?
1.02
81.
034
1.04
21.
052?
1.04
9??
1.01
61.
039?
1.04
01.
052?
1.04
4?1.
014
1.03
8?1.
042?
1.04
9?1.
043?
21.
011
1.00
71.
014
1.01
21.
032?
1.01
11.
008
1.01
61.
018
1.03
4?1.
005
1.01
21.
026
1.02
21.
030?
1.00
51.
012
1.02
41.
017
1.02
8?
31.
011
1.00
01.
019
0.99
61.
034?
1.01
01.
002
1.01
81.
002
1.03
7?0.
999
1.00
61.
041?
1.01
11.
033?
1.00
01.
007
1.03
7?1.
004
1.03
1?
41.
012
1.00
81.
031??
0.98
51.
037?
1.01
01.
007
1.02
9??
0.99
21.
040?
0.99
31.
006
1.05
6???
1.00
41.
035?
0.99
31.
009
1.04
8???
0.99
51.
033?
hM
achi
nery
and
Equi
pmen
t
11.
027
0.98
61.
019
1.02
6?0.
983
1.02
90.
985
1.01
21.
024
0.98
01.
050
0.99
61.
001
1.00
20.
976
1.04
30.
999
1.00
71.
008?
0.98
4
21.
019
0.97
70.
975
0.99
20.
944??
1.02
20.
985
0.97
50.
994
0.95
0??
1.06
10.
991
0.96
80.
962
0.93
7??
1.05
10.
985
0.96
90.
970
0.94
2??
31.
028
1.02
60.
999
0.98
90.
950?
1.03
41.
017
0.99
70.
985
0.94
4?1.
079
1.01
71.
001
0.94
40.
928??
1.06
71.
018
1.00
00.
954
0.93
9?
41.
036
0.94
40.
996
1.02
20.
959??
1.04
30.
941
0.99
71.
018
0.95
5?1.
106
0.97
20.
997
0.97
30.
947?
1.08
90.
976
0.99
70.
988
0.95
9??
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
26
Tabl
e4.
A8:
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
fact
or-a
ugm
ente
d(2
-ter
m,i
nle
vels
)AR
(1)m
odel
sac
cord
ing
toit
sre
lati
onsh
ipw
ith
the
busi
ness
cycl
e(*
)PCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEICCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
893
1.01
60.
903
0.99
21.
057
0.91
21.
114
0.90
90.
924
1.03
41.
119
1.03
71.
055
0.92
91.
007
1.02
60.
915
1.04
60.
905
0.92
7
20.
921
1.23
5?0.
906
0.90
41.
067
0.93
31.
251??
0.92
10.
909
1.07
00.
944
1.10
8??
0.99
81.
045
1.07
50.
913
1.06
30.
975
1.03
91.
058
30.
920
1.26
6???
0.88
40.
976
1.07
60.
908
1.26
4??
0.90
01.
006
1.08
90.
881?
1.16
5?1.
052
1.10
01.
096
0.86
6?1.
121?
1.03
81.
080
1.07
6
40.
906
1.27
5???
0.85
61.
016
1.05
50.
901
1.29
0???
0.87
31.
036
1.07
20.
888?
1.17
4??
1.07
21.
110
1.08
80.
866?
1.10
71.
056
1.07
41.
053
hD
urab
leC
onsu
mpt
ion
10.
996
0.96
40.
961
0.98
50.
957??
0.99
40.
974
0.96
50.
953??
0.94
5??
1.00
71.
016
1.02
20.
993
1.01
30.
992
1.00
90.
996
0.96
60.
995
20.
996
0.98
7?0.
981
0.99
00.
976
1.00
10.
993
0.98
30.
989
0.97
81.
025
1.00
31.
078
1.01
51.
008
1.01
71.
000
1.07
11.
006
1.00
7
30.
997
0.98
30.
980
1.01
00.
981
0.99
50.
980
0.98
71.
010
0.97
51.
025?
1.00
11.
100
1.00
70.
993
1.02
21.
001
1.09
80.
999
0.99
7
40.
993
0.98
10.
973
1.00
50.
969
0.99
20.
982
0.97
81.
015
0.96
91.
030
1.00
01.
094
1.01
50.
993
1.02
50.
997
1.10
11.
003
0.99
7
hN
on-d
urab
leC
onsu
mpt
ion
10.
888?
1.16
30.
957
0.94
91.
057
0.89
51.
186??
0.96
20.
911
1.01
71.
066
0.93
20.
993
0.90
5??
0.91
41.
001
0.84
5?1.
001
0.88
8?0.
872?
20.
897??
1.29
4???
0.91
60.
845?
0.98
70.
893?
1.29
3???
0.91
60.
857?
0.98
90.
883?
1.09
4???
0.94
50.
980
0.99
80.
870??
1.06
4?0.
931?
0.95
70.
989
30.
895??
1.27
3???
0.90
0?0.
870
0.97
80.
884??
1.27
7???
0.90
20.
882
0.98
60.
846???
1.12
3???
0.95
70.
984
0.99
90.
837???
1.07
40.
954
0.95
20.
979
40.
891??
1.30
6???
0.86
9?0.
890
0.98
40.
885??
1.31
4???
0.87
1?0.
893
0.98
90.
839???
1.16
7???
0.95
40.
995
1.01
60.
834???
1.11
8??
0.94
60.
959
0.99
5
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
028
0.98
90.
999
1.01
70.
990
1.02
3?0.
994
0.99
21.
021
0.99
81.
033
1.01
10.
982
0.98
00.
996
1.03
11.
011
0.98
60.
990
1.00
3
20.
997
0.97
60.
966
1.00
70.
992
0.99
80.
989
0.96
81.
025
1.01
01.
029
1.00
80.
961??
0.97
00.
992
1.02
21.
002
0.95
1???
0.97
10.
989
31.
005
1.01
70.
992
1.00
61.
021
1.00
71.
013
0.99
41.
019
1.02
81.
017
1.02
40.
992
0.99
31.
017
1.01
71.
022
0.98
50.
985
1.01
4
41.
000
1.00
90.
988
0.99
51.
014
1.00
31.
006
0.98
81.
010
1.02
11.
018
1.01
70.
994
0.98
1?1.
009
1.01
41.
023
0.98
5??
0.97
5?1.
011
hC
onst
ruct
ion
and
Wor
ks
11.
019
1.01
21.
005
1.00
81.
011
1.01
91.
014
1.00
61.
020
1.01
81.
001
1.02
21.
003
1.02
31.
010
0.99
61.
018
1.00
91.
017
1.00
4
21.
008
1.01
20.
985
0.99
21.
007
1.00
81.
012
0.98
41.
000
1.01
11.
003
1.02
50.
999
1.01
11.
006
1.00
11.
029
0.99
71.
001
1.00
0
31.
010
1.01
60.
980
0.98
31.
010
1.01
01.
018
0.97
50.
991
1.01
30.
998
1.03
1?1.
009
1.00
91.
010
0.99
71.
039
1.00
30.
995
1.00
5
41.
014
1.02
30.
977
0.97
51.
013
1.01
31.
022
0.97
00.
984
1.01
60.
988
1.02
21.
010
1.00
91.
010
0.98
71.
030
0.99
80.
991
1.00
5
hM
achi
nery
and
Equi
pmen
t
11.
008
0.99
81.
015
1.03
11.
019
1.01
20.
994
1.00
61.
027
1.01
21.
055
1.00
30.
974
0.98
81.
001
1.04
71.
011
0.97
50.
998
1.01
5
20.
995
0.97
10.
987
1.03
81.
020
0.99
60.
981
0.98
51.
046
1.02
9?1.
054
0.99
40.
967
0.98
81.
006
1.04
50.
991
0.96
11.
001
1.01
7
31.
003
1.00
01.
003
1.10
51.
089?
1.01
10.
988
0.99
61.
102
1.07
71.
087
0.98
00.
990
1.00
81.
025
1.07
50.
980
0.97
71.
017
1.03
3
40.
994
0.96
70.
983
1.08
91.
067
1.00
10.
960
0.98
21.
091
1.05
81.
110
1.01
40.
970
1.00
21.
032
1.09
11.
030
0.95
91.
026
1.05
5
(*)R
elat
ive
RM
SFE2 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=
Figu
res
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Har
vey,
Leyb
ourn
e,an
dN
ewbo
ld(1
997)
test
’sp-
valu
e:(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
27
9R
obus
tnes
sre
sult
sIX
:AR
(1)n
ativ
em
odel
,3-t
erm
fact
orin
leve
ls
Tabl
e1.
A9.
Adj
uste
dR
-squ
ared
rati
obe
twee
nA
R(1
)fac
tor-
augm
ente
d(3
-ter
m,i
nle
vels
)and
nati
veA
R(1
)mod
el(*
)an
dF-
test
resu
lts
(p-v
alue
)oft
heco
nsum
er-b
ased
fact
orst
atis
tica
lsig
nific
ance
(**)
Fact
or:M
ore
sens
itiv
ese
ctor
sFa
ctor
:Les
sse
nsit
ive
sect
ors
Fact
or:M
ore
sens
itiv
ese
ctor
sFa
ctor
:Les
sse
nsit
ive
sect
ors
PCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEI
Fact
or:C
omm
erce
+In
dust
ryPC
1.01
21.
030
1.05
01.
032
1.03
90.
996
1.01
31.
078
1.03
81.
042
0.46
70.
160
0.00
30.
040
0.02
00.
822
0.46
50.
000
0.02
10.
021
DC
1.00
51.
042
1.05
01.
047
1.05
70.
991
1.00
81.
062
1.04
01.
043
0.29
40.
006
0.00
20.
018
0.00
30.
855
0.26
70.
001
0.01
50.
007
ND
C0.
997
1.04
61.
031
1.02
71.
042
0.98
41.
030
1.06
11.
034
1.04
70.
555
0.22
70.
060
0.09
70.
035
0.68
70.
371
0.01
40.
059
0.03
8G
FCF
1.00
01.
007
1.03
81.
008
1.01
30.
994
0.99
41.
070
1.02
31.
022
0.45
70.
417
0.02
30.
262
0.15
40.
687
0.68
70.
001
0.10
70.
106
CW
K0.
984
0.99
20.
981
0.99
80.
999
0.98
80.
984
0.98
80.
990
0.98
80.
894
0.68
70.
984
0.60
80.
575
0.72
40.
923
0.72
80.
692
0.82
5M
EQ1.
006
1.02
71.
113
1.05
11.
056
0.98
71.
020
1.17
01.
081
1.08
70.
517
0.21
80.
001
0.04
10.
016
0.89
20.
205
0.00
00.
024
0.01
0Fa
ctor
:Com
mer
ce+
Indu
stry
+Tr
ansp
orta
tion
PC1.
007
1.02
61.
058
1.03
41.
040
0.99
81.
015
1.07
51.
038
1.04
10.
538
0.23
90.
001
0.04
40.
025
0.79
10.
409
0.00
00.
019
0.01
8D
C1.
001
1.03
91.
054
1.04
51.
055
0.99
21.
007
1.06
11.
039
1.04
30.
472
0.00
80.
002
0.02
30.
004
0.80
80.
282
0.00
10.
013
0.00
6N
DC
0.99
21.
044
1.03
61.
026
1.04
20.
988
1.03
21.
061
1.03
61.
048
0.63
00.
372
0.05
60.
123
0.04
50.
584
0.27
50.
012
0.04
30.
032
GFC
F0.
997
1.00
21.
043
1.00
71.
011
0.99
40.
995
1.07
11.
025
1.02
50.
529
0.50
30.
013
0.29
20.
198
0.73
90.
698
0.00
10.
087
0.08
6C
WK
0.98
30.
989
0.98
20.
995
0.99
50.
987
0.98
50.
987
0.99
20.
989
0.90
30.
766
0.95
40.
691
0.67
10.
770
0.90
40.
720
0.61
40.
768
MEQ
1.00
11.
026
1.11
61.
048
1.05
40.
987
1.01
81.
177
1.08
71.
092
0.53
70.
223
0.00
10.
054
0.02
00.
892
0.24
30.
000
0.02
10.
008
Fact
or:C
omm
erce
+In
dust
ry+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
994
1.00
51.
078
1.03
81.
039
1.01
31.
043
1.05
21.
034
1.04
30.
882
0.61
80.
000
0.02
10.
029
0.42
70.
072
0.00
40.
038
0.01
3D
C0.
987
1.00
11.
060
1.03
91.
041
1.00
81.
048
1.05
41.
047
1.05
70.
915
0.43
40.
001
0.01
60.
007
0.30
60.
003
0.00
10.
020
0.00
3N
DC
0.98
21.
017
1.06
11.
037
1.04
50.
999
1.06
81.
031
1.02
51.
046
0.70
10.
459
0.01
40.
048
0.05
00.
467
0.11
00.
071
0.11
20.
026
GFC
F0.
991
0.99
31.
068
1.02
51.
023
1.00
21.
006
1.04
71.
008
1.01
30.
754
0.73
20.
001
0.08
20.
087
0.37
90.
460
0.00
70.
278
0.17
9C
WK
0.98
70.
991
0.98
50.
992
0.99
00.
988
0.98
40.
983
0.99
50.
994
0.72
90.
734
0.82
00.
614
0.76
30.
784
0.91
50.
917
0.68
60.
696
MEQ
0.98
41.
019
1.16
81.
087
1.09
21.
005
1.02
51.
125
1.05
01.
056
0.94
10.
204
0.00
00.
014
0.00
60.
506
0.24
10.
000
0.06
10.
025
Fact
or:C
omm
erce
+C
onst
ruct
ion
+Pe
rson
alSe
rvic
esPC
0.99
71.
010
1.07
51.
035
1.03
71.
010
1.05
01.
051
1.03
81.
048
0.81
20.
475
0.00
00.
029
0.03
50.
495
0.06
90.
003
0.02
40.
007
DC
0.99
11.
007
1.05
71.
039
1.04
21.
005
1.06
21.
059
1.05
01.
063
0.84
50.
311
0.00
20.
022
0.01
10.
297
0.00
10.
001
0.01
30.
001
ND
C0.
983
1.02
51.
054
1.03
51.
046
0.99
91.
077
1.03
61.
026
1.04
60.
739
0.35
20.
026
0.05
70.
048
0.46
00.
122
0.04
00.
120
0.02
8G
FCF
0.99
20.
995
1.06
41.
023
1.02
21.
008
1.00
41.
046
1.00
41.
009
0.76
40.
687
0.00
20.
095
0.09
30.
310
0.49
00.
006
0.35
10.
232
CW
K0.
985
0.99
10.
984
0.99
40.
992
0.99
00.
985
0.98
20.
992
0.99
10.
822
0.70
70.
874
0.59
40.
688
0.73
90.
861
0.89
70.
761
0.76
1M
EQ0.
986
1.02
51.
166
1.08
71.
091
1.01
51.
013
1.11
51.
036
1.04
10.
920
0.20
00.
000
0.01
30.
005
0.40
20.
365
0.00
00.
103
0.05
2(*
)Adj
uste
dR
-squ
ared
rati
o=A
djus
ted
R-s
quar
edw
ith
the
fact
or/A
djus
ted
R-s
quar
edw
itho
utth
efa
ctor
.Ora
nge-
shad
edce
lls=A
djus
ted
R-s
quar
edra
tio
grea
ter
than
1.G
reen
-sha
ded
cells
=Adj
uste
dR
-squ
ared
rati
ogr
eate
rth
an1.
05(5
%of
adju
stm
entg
ain)
.Sam
ple:
2001
.I-20
14.IV
(56
obse
rvat
ions
).(*
*)p-
valu
eof
the
null
hypo
thes
is:N
H:φ
1=...
=φ
4=0
(see
equa
tion
1)ag
ains
tthe
alte
rnat
ive
ofat
leas
tone
term
iseq
ual
toze
ro.G
reen
-sha
ded
cells
=p<
10%
.Ora
nge-
shad
edce
lls=p<
15%
.Sam
ple:
2001
.I-20
19.II
I(75
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
28
Tabl
e2.
A9.
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(1
)mod
el(3
-ter
m,i
nle
vels
)(*)
PCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
974
0.88
31.
196???
1.07
90.
957
1.00
00.
916
1.22
9??
1.06
10.
946
1.10
60.
874
1.23
9???
1.05
50.
941
1.06
50.
848
1.21
7???
1.05
20.
932
21.
033?
0.99
91.
341??
1.10
81.
024
1.02
40.
976
1.38
0??
1.14
81.
030
1.08
50.
890
1.36
7??
1.23
0?1.
054
1.04
70.
902
1.34
9???
1.21
2?1.
058
30.
899
1.05
01.
038
1.04
70.
999
0.88
11.
034
1.07
61.
085
1.01
70.
955
0.98
51.
199??
1.14
61.
043
0.92
40.
992
1.17
9??
1.12
61.
040
40.
832
1.10
20.
950
1.10
3??
1.02
00.
831
1.09
60.
977
1.11
9??
1.03
40.
886
1.04
71.
123??
1.18
5??
1.07
40.
853
1.04
81.
101??
1.16
1??
1.06
3
hD
urab
leC
onsu
mpt
ion
10.
999
0.97
30.
954
0.93
40.
917??
1.00
10.
980
0.96
00.
913
0.90
9??
1.01
31.
001
0.98
40.
946
0.94
81.
007
0.99
80.
971
0.93
70.
943
21.
015
1.00
21.
018
1.00
00.
962
1.02
01.
008
1.02
41.
006
0.96
51.
036??
1.01
21.
062
1.03
60.
986
1.03
4??
1.01
11.
055
1.03
30.
985
31.
023
0.99
40.
976
1.04
60.
971
1.02
30.
995
0.98
31.
045
0.96
81.
046?
1.00
61.
044
1.03
70.
980
1.04
4?1.
005
1.03
91.
037
0.98
1
41.
024
0.98
30.
954
1.07
20.
963
1.02
70.
986
0.95
81.
076
0.96
41.
053
0.99
71.
022
1.05
70.
976
1.05
00.
994
1.01
81.
057
0.97
6
hN
on-d
urab
leC
onsu
mpt
ion
10.
978
0.86
91.
189??
1.10
31.
025
0.99
10.
861
1.20
8??
1.08
51.
005
1.09
20.
748??
1.19
7???
1.06
70.
967
1.06
40.
741??
1.18
2???
1.06
30.
965
21.
007
0.95
91.
229
1.07
61.
042
0.99
30.
933
1.24
51.
108
1.05
41.
062
0.79
91.
236?
1.16
11.
063
1.03
50.
823
1.22
0?1.
133
1.05
8
30.
891
1.08
10.
970
0.99
81.
026
0.87
91.
071
0.98
11.
015
1.03
90.
943
0.98
31.
044
1.07
01.
062
0.92
10.
988
1.03
21.
039
1.05
0
40.
817???
1.15
2???
0.92
51.
003
1.05
40.
815???
1.14
2???
0.93
11.
008
1.06
30.
847
1.07
10.
995
1.09
01.
103
0.83
11.
081?
0.98
31.
056
1.09
1
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
029
0.97
31.
014
1.05
20.
998
1.03
60.
991
1.01
11.
056
1.00
81.
056
1.01
20.
984
1.02
30.
999
1.05
00.
999
0.98
91.
031
0.99
9
21.
009
0.95
91.
011
1.03
50.
984
1.01
70.
975
1.01
91.
042
0.99
91.
056
0.98
60.
988
1.00
40.
978
1.04
20.
968
0.98
71.
008
0.97
2
31.
027
1.00
11.
028
1.00
71.
008
1.02
30.
999
1.03
31.
007
1.01
01.
018
1.00
71.
036
0.98
50.
995
1.01
81.
001
1.03
30.
986
0.99
2
41.
023
0.98
61.
016
0.99
10.
995
1.02
20.
986
1.01
40.
990
0.99
81.
024
0.99
11.
012
0.96
20.
980
1.02
20.
988
1.00
90.
965
0.98
0
hC
onst
ruct
ion
and
Wor
ks
11.
016
1.04
5?1.
073?
1.08
1?1.
068??
1.01
51.
046?
1.08
0?1.
093??
1.07
5??
1.00
51.
052
1.07
5??
1.09
3??
1.07
2?1.
002
1.04
91.
074??
1.08
8?1.
068?
21.
015
1.01
21.
054?
1.05
91.
055??
1.01
11.
013
1.05
8??
1.07
3?1.
062??
0.99
81.
024
1.06
2??
1.08
4?1.
067??
0.99
71.
022
1.05
8?1.
076?
1.06
1?
31.
022
0.99
81.
047?
1.02
01.
049??
1.01
81.
000
1.04
5??
1.02
91.
052??
0.99
31.
007
1.06
6???
1.04
91.
054??
0.99
41.
006
1.06
0???
1.04
01.
051??
41.
024
1.00
41.
055??
0.98
81.
044?
1.01
81.
003
1.05
1??
0.99
81.
046?
0.99
21.
001
1.07
4???
1.02
11.
042?
0.99
41.
002
1.06
4???
1.01
01.
040?
hM
achi
nery
and
Equi
pmen
t
11.
025
0.98
81.
013
1.04
2??
0.98
61.
030?
0.99
51.
009
1.03
6?0.
987
1.03
1?1.
011
0.99
11.
019
0.98
41.
033?
1.00
80.
994
1.02
5?0.
989
21.
018
0.98
60.
981
1.01
50.
941??
1.02
40.
991
0.98
61.
009
0.94
9?1.
058
1.00
00.
966
0.97
50.
933
1.05
00.
993
0.96
40.
981
0.93
4
31.
021
1.03
90.
973
1.01
80.
950
1.02
21.
025
0.98
50.
997
0.94
31.
012
1.03
50.
988
0.94
70.
922
1.01
21.
036
0.98
60.
957
0.92
7
41.
029
0.95
50.
966
1.04
10.
967
1.03
30.
954
0.96
61.
030
0.96
3?1.
040
0.99
50.
969
0.97
80.
952
1.03
70.
996
0.97
60.
997
0.96
2
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
29
Tabl
e4.
A9:
Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
fact
or-a
ugm
ente
d(3
-ter
m,i
nle
vels
)AR
(1)m
odel
sac
cord
ing
toit
sre
lati
onsh
ipw
ith
the
busi
ness
cycl
e(*
)PCSI
PFEI
CCSICFEIOFEIPCSI
PFEI
CCSICFEIOFEIPCSI
PFEI
CCSI
CFEIOFEI
PCSI
PFEICCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
967
1.02
41.
119
1.19
8??
1.15
3?0.
996
1.09
91.
156
1.14
3?1.
130
1.24
1?1.
011
1.20
41.
107
1.09
11.
142
0.92
51.
178
1.07
91.
027
21.
002
1.19
11.
143
0.99
61.
062
0.99
81.
170
1.19
11.
034
1.07
01.
059
0.99
71.
169
1.15
21.
093
1.00
20.
991
1.14
7?1.
135
1.10
0
30.
894
1.20
3?0.
929
0.99
81.
037
0.87
21.
184
0.96
51.
045
1.06
50.
927
1.08
01.
125
1.10
11.
071
0.88
71.
052
1.11
21.
074
1.06
0
40.
845
1.27
6??
0.89
51.
102
1.07
10.
847
1.28
2??
0.91
91.
124
1.09
60.
888
1.14
9?1.
125
1.17
9?1.
114
0.84
21.
083
1.10
01.
130
1.07
4
hD
urab
leC
onsu
mpt
ion
11.
003
0.97
60.
989
0.98
40.
956
1.00
30.
983
0.99
60.
945
0.93
81.
016
1.01
91.
037
0.99
51.
004
1.00
21.
012
1.00
80.
966
0.98
8
21.
005
0.99
51.
008
0.95
30.
957
1.00
91.
003
1.01
20.
953
0.95
81.
034
1.01
11.
079
1.01
70.
995
1.02
81.
011
1.07
21.
013
0.99
8
31.
007
0.98
50.
990
1.02
80.
982
1.00
40.
986
0.99
71.
022
0.97
51.
041?
1.00
11.
104
1.01
50.
993
1.03
8?1.
002
1.10
31.
010
0.99
7
41.
005
0.98
20.
984
1.06
80.
993
1.00
60.
987
0.98
51.
074
0.99
31.
051
1.00
31.
097
1.04
11.
011
1.04
41.
003
1.10
11.
030
1.01
4
hN
on-d
urab
leC
onsu
mpt
ion
10.
953
1.14
3??
1.09
81.
103
1.12
6??
0.96
11.
146??
1.12
61.
060
1.08
8?1.
186
0.87
4?1.
116
1.01
10.
980
1.12
30.
831??
1.10
80.
991
0.94
8
20.
961
1.28
7??
1.07
90.
972
1.02
30.
943
1.25
0??
1.10
21.
009
1.04
11.
023
0.94
21.
075
1.07
81.
031
0.99
00.
943
1.05
31.
035
1.01
5
30.
866??
1.28
9???
0.91
80.
951
0.99
30.
851???
1.28
5???
0.93
00.
972
1.01
20.
912
1.07
11.
002
1.01
61.
009
0.88
4?1.
016
0.99
50.
966
0.97
6
40.
820???
1.38
5???
0.89
10.
974
1.01
80.
818???
1.38
4???
0.89
50.
978
1.03
10.
840?
1.16
6???
0.97
81.
061
1.05
10.
822??
1.11
6?0.
964
1.00
11.
017
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
009
0.97
91.
028
1.03
51.
004
1.01
61.
003
1.03
21.
043
1.02
11.
070
1.02
90.
974
0.98
21.
005
1.05
00.
991
0.96
80.
993
0.99
5
20.
993
0.97
01.
022
1.03
61.
016
1.00
10.
991
1.04
01.
055??
1.04
0?1.
066
1.01
40.
977
0.97
11.
004
1.04
20.
980
0.96
40.
969
0.97
7
31.
017
1.01
51.
025
1.04
8?1.
061??
1.01
31.
012
1.03
41.
051?
1.06
6?1.
007
1.02
41.
028
0.97
61.
026
1.00
71.
005
1.01
90.
962
0.99
8
41.
013
1.00
31.
031
1.03
7?1.
049?
1.01
21.
002
1.03
11.
039??
1.05
51.
020
1.01
31.
009
0.95
91.
012
1.01
61.
010
0.99
80.
952
0.99
6
hC
onst
ruct
ion
and
Wor
ks
11.
004
1.00
71.
013
1.00
81.
003
1.00
31.
009
1.02
01.
028
1.01
40.
984
1.00
91.
013
1.03
51.
008
0.97
50.
994
1.01
61.
022
0.99
3
21.
010
0.99
11.
001
0.98
00.
986
1.00
60.
995
1.00
30.
998
0.99
40.
990
1.00
81.
014
1.02
61.
007
0.98
71.
006
1.01
01.
006
0.99
0
31.
020
0.98
80.
992
0.95
90.
985
1.01
80.
992
0.98
70.
970
0.98
80.
985
1.00
61.
023
1.01
10.
998
0.98
61.
007
1.01
20.
989
0.98
6
41.
026
0.99
60.
992
0.94
70.
986
1.02
10.
997
0.98
40.
960
0.99
00.
983
0.99
91.
024
1.00
40.
992
0.98
41.
001
1.00
50.
980
0.98
1
hM
achi
nery
and
Equi
pmen
t
11.
004
0.99
11.
043
1.03
3?1.
018
1.00
91.
001
1.04
21.
023
1.01
71.
042?
1.01
91.
001
0.99
01.
009
1.03
30.
998
0.99
30.
998
1.00
8
20.
996
0.95
71.
016
1.04
51.
013
1.00
00.
963
1.02
71.
044
1.02
41.
059
0.98
10.
994
0.98
40.
998
1.04
50.
981
0.97
60.
992
0.99
2
31.
006
0.98
01.
067
1.13
5??
1.09
11.
011
0.96
01.
087
1.09
9?1.
076
1.00
30.
977
1.10
6??
0.97
91.
015
1.00
40.
985
1.10
7?0.
981
1.00
1
40.
997
0.93
21.
052
1.10
41.
055
1.00
30.
925
1.05
21.
090
1.04
71.
028
1.00
61.
047
0.97
91.
018
1.02
21.
023
1.06
61.
006
1.02
9
(*)R
elat
ive
RM
SFE2 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=
Figu
res
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Har
vey,
Leyb
ourn
e,an
dN
ewbo
ld(1
997)
test
’sp-
valu
e:(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
30
10R
obus
tnes
sre
sult
sX
:AR
(1)n
ativ
em
odel
,4-t
erm
fact
orin
leve
ls
Tabl
e1.
A10
Adj
uste
dR
-squ
ared
rati
obe
twee
nA
R(1
)fac
tor-
augm
ente
d(4
-ter
m,i
nle
vels
)and
nati
veA
R(1
)mod
el(*
)an
dF-
test
resu
lts
(p-v
alue
)oft
heco
nsum
er-b
ased
fact
orst
atis
tica
lsig
nific
ance
(**)
Fact
or:M
ore
sens
itiv
ese
ctor
sFa
ctor
:Les
sse
nsit
ive
sect
ors
Fact
or:M
ore
sens
itiv
ese
ctor
sFa
ctor
:Les
sse
nsit
ive
sect
ors
PCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEIPCSIPFEICCSICFEIOFEI
Fact
or:C
omm
erce
+In
dust
ryPC
1.01
41.
037
1.05
41.
039
1.04
80.
990
1.01
61.
074
1.03
91.
045
0.21
50.
144
0.00
10.
035
0.01
00.
846
0.19
90.
000
0.03
20.
017
DC
1.00
71.
059
1.06
61.
076
1.08
80.
981
1.01
51.
061
1.05
71.
062
0.20
10.
002
0.00
00.
000
0.00
00.
941
0.02
90.
001
0.00
10.
000
ND
C0.
979
1.03
41.
031
1.02
51.
040
0.96
61.
018
1.05
01.
029
1.04
20.
710
0.34
80.
062
0.11
30.
034
0.80
60.
547
0.01
60.
091
0.05
6G
FCF
1.00
51.
001
1.03
21.
001
1.00
60.
995
0.99
01.
065
1.01
71.
016
0.27
90.
566
0.05
50.
407
0.27
60.
537
0.69
60.
003
0.18
90.
166
CW
K0.
977
0.98
90.
974
0.99
10.
993
0.97
90.
979
0.98
40.
985
0.98
30.
918
0.68
80.
899
0.75
00.
688
0.81
70.
946
0.35
00.
707
0.76
3M
EQ1.
011
1.02
01.
102
1.04
41.
047
0.98
61.
014
1.16
01.
074
1.07
90.
425
0.31
20.
004
0.06
50.
032
0.80
10.
289
0.00
00.
035
0.01
6Fa
ctor
:Com
mer
ce+
Indu
stry
+Tr
ansp
orta
tion
PC1.
006
1.03
51.
058
1.03
81.
047
0.99
21.
018
1.07
11.
039
1.04
40.
459
0.20
00.
001
0.04
40.
016
0.78
50.
178
0.00
00.
026
0.01
4D
C0.
998
1.05
81.
061
1.07
21.
084
0.98
31.
014
1.06
21.
058
1.06
20.
528
0.00
10.
001
0.00
00.
000
0.89
80.
038
0.00
00.
001
0.00
0N
DC
0.97
41.
032
1.03
31.
026
1.04
10.
969
1.02
01.
049
1.02
91.
041
0.78
40.
501
0.06
50.
131
0.03
70.
733
0.41
30.
014
0.06
90.
048
GFC
F0.
996
0.99
71.
036
1.00
11.
004
0.99
90.
991
1.06
81.
018
1.01
80.
481
0.64
20.
034
0.42
50.
324
0.41
00.
659
0.00
30.
166
0.13
6C
WK
0.97
40.
986
0.97
50.
987
0.98
80.
980
0.98
00.
984
0.98
70.
987
0.95
60.
784
0.86
40.
826
0.78
10.
798
0.94
10.
301
0.58
60.
655
MEQ
1.00
01.
020
1.10
51.
042
1.04
80.
991
1.01
31.
168
1.07
91.
083
0.55
50.
320
0.00
30.
066
0.03
30.
710
0.30
80.
000
0.03
70.
016
Fact
or:C
omm
erce
+In
dust
ry+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
995
1.01
31.
075
1.03
91.
044
1.00
71.
043
1.05
21.
038
1.04
80.
621
0.18
20.
000
0.03
10.
015
0.42
00.
114
0.00
20.
039
0.01
2D
C0.
980
1.01
21.
060
1.05
71.
063
1.00
21.
058
1.06
31.
072
1.08
20.
941
0.05
5.0
010.
001
0.00
00.
369
0.00
00.
001
0.00
00.
000
ND
C0.
969
1.01
01.
051
1.03
11.
042
0.98
11.
053
1.02
31.
024
1.04
00.
706
0.53
00.
015
0.07
60.
060
0.67
30.
219
0.07
30.
128
0.03
3G
FCF
0.99
30.
988
1.06
31.
018
1.01
61.
005
1.00
21.
041
1.00
21.
007
0.63
10.
765
0.00
40.
169
0.15
40.
242
0.55
60.
020
0.37
10.
273
CW
K0.
978
0.98
40.
981
0.98
50.
985
0.98
20.
981
0.97
80.
989
0.98
90.
848
0.87
40.
420
0.65
20.
735
0.81
70.
852
0.70
00.
804
0.77
0M
EQ0.
987
1.01
31.
159
1.07
91.
083
1.00
41.
019
1.11
51.
043
1.04
90.
785
0.28
90.
000
0.02
50.
011
0.51
50.
318
0.00
20.
075
0.03
9Fa
ctor
:Com
mer
ce+
Con
stru
ctio
n+
Pers
onal
Serv
ices
PC0.
997
1.01
81.
071
1.03
81.
044
1.00
21.
050
1.05
41.
040
1.05
00.
645
0.17
60.
000
0.03
50.
016
0.56
80.
059
0.00
10.
029
0.00
8D
C0.
984
1.02
11.
057
1.06
01.
067
0.99
81.
069
1.06
91.
072
1.08
20.
874
0.04
40.
001
0.00
10.
000
0.41
20.
000
0.00
10.
000
0.00
0N
DC
0.96
81.
016
1.04
21.
031
1.04
30.
983
1.06
21.
030
1.02
31.
040
0.78
80.
450
0.03
20.
081
0.05
40.
677
0.22
00.
032
0.14
20.
042
GFC
F0.
991
0.99
01.
058
1.01
61.
016
1.01
61.
004
1.04
00.
999
1.00
40.
675
0.79
10.
006
0.18
30.
168
0.11
40.
384
0.01
50.
445
0.29
7C
WK
0.97
60.
985
0.98
00.
988
0.98
80.
986
0.98
10.
979
0.98
40.
984
0.91
70.
837
0.50
90.
649
0.69
00.
731
0.84
00.
662
0.87
30.
861
MEQ
0.98
51.
017
1.15
61.
079
1.08
21.
017
1.01
31.
105
1.03
11.
036
0.82
40.
322
0.00
00.
022
0.01
20.
381
0.28
40.
002
0.11
00.
055
(*)A
djus
ted
R-s
quar
edra
tio=
Adj
uste
dR
-squ
ared
wit
hth
efa
ctor
/Adj
uste
dR
-squ
ared
wit
hout
the
fact
or.O
rang
e-sh
aded
cells
=Adj
uste
dR
-squ
ared
rati
ogr
eate
rth
an1.
Gre
en-s
hade
dce
lls=A
djus
ted
R-s
quar
edra
tio
grea
ter
than
1.05
(5%
ofad
just
men
tgai
n).S
ampl
e:20
01.I-
2014
.IV(5
6ob
serv
atio
ns).
(**)
p-va
lue
ofth
enu
llhy
poth
esis
:NH
:φ1=
...=
φ4=
0(s
eeeq
uati
on1)
agai
nstt
heal
tern
ativ
eof
atle
asto
nete
rmis
equa
lto
zero
.Gre
en-s
hade
dce
lls=p<
10%
.Ora
nge-
shad
edce
lls=p<
15%
.Sam
ple:
2001
.I-20
19.II
I(75
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
31
Tabl
e2.
A10
.Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
nati
veA
R(1
)and
fact
or-a
ugm
ente
dA
R(1
)mod
el(4
-ter
m,i
nle
vels
)(*)
PCSI
PFEICCSI
CFEIOFEIPCSI
PFEICCSI
CFEIOFEIPCSI
PFEI
CCSI
CFEI
OFEIPCSI
PFEI
CCSI
CFEI
OFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
960
0.96
41.
223??
1.23
4?1.
046?
0.97
50.
988
1.25
5??
1.23
71.
047?
1.16
0??
0.99
71.
320???
1.21
81.
062?
1.11
0?0.
977
1.30
8???
1.21
81.
055?
21.
082??
0.97
41.
534???
1.45
4?1.
114?
1.06
9??
0.95
31.
594???
1.48
6?1.
130?
1.27
3??
0.93
71.
543???
1.46
8?1.
153??
1.20
6??
0.93
21.
529???
1.46
0?1.
151??
30.
955
0.92
41.
314???
1.37
9??
1.03
70.
922
0.89
91.
360???
1.39
9??
1.05
91.
140??
0.93
91.
372???
1.38
6??
1.10
2?1.
063?
0.93
71.
360???
1.37
6??
1.09
8?
40.
754
0.95
51.
025??
1.21
1???
0.98
50.
761
0.94
11.
054???
1.19
7???
0.99
60.
983??
0.90
31.
166???
1.24
6???
1.03
6?0.
914?
0.90
11.
137???
1.22
0???
1.02
3?
hD
urab
leC
onsu
mpt
ion
11.
007
0.98
90.
951
0.94
40.
932
1.00
60.
995
0.96
60.
928
0.92
71.
022
1.01
70.
994
0.96
20.
968
1.01
51.
015
0.98
50.
955
0.96
3
21.
030?
1.03
31.
043
1.01
20.
988
1.03
7??
1.03
11.
054?
1.01
70.
988
1.06
3??
1.02
71.
064?
1.05
11.
006
1.05
7??
1.02
81.
059?
1.04
81.
005
31.
041
1.02
71.
002
1.05
10.
994
1.04
31.
022
1.01
31.
051
0.99
01.
071?
1.02
81.
032
1.04
51.
000
1.06
6?1.
030
1.03
21.
048
1.00
1
41.
043
1.00
30.
985
1.06
8?0.
977
1.04
81.
004
0.98
91.
071?
0.97
71.
084?
1.00
81.
012
1.05
0?0.
980
1.07
8?1.
006
1.01
31.
054?
0.98
0
hN
on-d
urab
leC
onsu
mpt
ion
10.
998
0.85
91.
160??
1.24
6?1.
111
1.00
90.
862
1.18
2???
1.24
2?1.
100
1.19
80.
788
1.22
1???
1.20
71.
065
1.15
40.
769
1.21
9???
1.20
71.
060
21.
068?
0.73
71.
317??
1.40
6?1.
163
1.05
6??
0.73
11.
359???
1.42
0?1.
171
1.29
80.
656
1.36
3???
1.39
71.
160
1.24
50.
653
1.34
6???
1.37
4?1.
149
30.
923
0.82
41.
143???
1.33
5?1.
138
0.90
90.
821
1.17
0???
1.33
3?1.
143
1.14
90.
753
1.20
2???
1.34
41.
134
1.09
20.
742
1.19
4???
1.31
21.
120
40.
698?
0.95
50.
933
1.15
9?1.
086
0.70
00.
945
0.94
71.
142?
1.08
90.
901
0.86
71.
060
1.21
81.
112
0.85
70.
873
1.03
51.
171
1.09
6
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
040
0.99
11.
040
1.07
0?1.
017
1.04
51.
012
1.03
71.
073
1.02
71.
059
1.04
00.
994
1.04
71.
031
1.05
71.
029
1.00
11.
056
1.03
2
21.
013
0.98
31.
029
1.06
01.
014
1.02
41.
002
1.03
91.
068
1.02
41.
062
1.02
50.
998
1.05
51.
032
1.04
91.
011
0.99
71.
060
1.02
9
31.
053?
1.03
11.
050?
1.04
3?1.
044?
1.05
01.
029
1.05
7?1.
047?
1.04
11.
046
1.05
31.
062??
1.08
4??
1.07
1??
1.04
31.
055
1.06
3??
1.08
0??
1.07
3??
41.
034?
0.99
41.
018
1.02
11.
016
1.03
6?0.
995
1.01
81.
023?
1.01
51.
058?
1.01
41.
019
1.04
1??
1.03
1??
1.04
41.
015
1.02
01.
041??
1.03
5??
hC
onst
ruct
ion
and
Wor
ks
11.
019
1.06
8?1.
109??
1.10
3?1.
086?
1.01
31.
067
1.11
5??
1.11
3?1.
091?
0.95
11.
071
1.10
1??
1.11
9??
1.09
4?0.
952
1.06
91.
098??
1.11
3?1.
089?
21.
016
1.03
41.
080???
1.08
41.
069
1.00
71.
033
1.08
6???
1.09
41.
072?
0.95
01.
037
1.08
9???
1.10
4?1.
080?
0.95
41.
037
1.08
3???
1.09
5?1.
073?
31.
034
1.02
11.
078??
1.02
21.
052
1.02
61.
019
1.07
6??
1.03
31.
053
0.96
31.
019
1.08
9???
1.06
11.
060
0.97
01.
021
1.08
1???
1.04
81.
055
41.
039
1.01
21.
073??
0.95
51.
033
1.03
31.
008
1.07
0???
0.96
91.
035
0.99
71.
000
1.08
2???
1.00
71.
030
0.99
91.
004
1.07
0???
0.99
21.
027
hM
achi
nery
and
Equi
pmen
t
11.
037?
0.99
61.
039
1.05
7??
1.00
51.
041??
1.01
11.
031
1.05
5??
1.00
91.
043??
1.02
40.
993
1.04
81.
011
1.04
4??
1.02
00.
998
1.05
41.
015
21.
026?
0.98
21.
006
1.06
4???
0.98
31.
032?
0.99
71.
011
1.05
9??
0.98
91.
068?
1.01
00.
968
1.03
70.
985
1.06
0?1.
000
0.96
91.
052?
0.99
0
31.
052
1.03
11.
019
1.13
3???
1.04
0?1.
052
1.03
21.
032?
1.11
9???
1.03
31.
034
1.07
2???
1.01
71.
116???
1.05
41.
038
1.07
2???
1.02
6?1.
138???
1.06
5?
41.
053
0.96
60.
967
1.09
4??
1.01
61.
056
0.97
10.
970
1.08
9??
1.01
21.
050?
1.02
20.
972
1.07
5??
1.02
51.
050?
1.02
10.
983
1.09
7??
1.03
5
(*)R
elat
ive
RM
SFE1 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=F
igur
esbe
low
1an
dst
atis
tica
llysi
gnifi
cant
.C
lark
and
Wes
t(20
07)t
est’s
p-va
lue:
(???
)p<
1%,
(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
32
Tabl
e4.
A10
:Mul
ti-h
oriz
onR
MSF
Era
tio
betw
een
fact
or-a
ugm
ente
d(4
-ter
m,i
nle
vels
)AR
(1)m
odel
sac
cord
ing
toit
sre
lati
onsh
ipw
ith
the
busi
ness
cycl
e(*
)PCSI
PFEI
CCSICFEIOFEIPCSI
PFEI
CCSICFEIOFEIPCSI
PFEICCSI
CFEIOFEI
PCSI
PFEICCSI
CFEIOFEI
Com
mer
ce+
Indu
stry
Com
mer
ce+
Indu
stry
+Tr
ansp
.C
omm
.+In
d.+
Con
st.+
Pers
.Ser
v.C
omm
.+C
onst
.+Pe
rs.S
erv.
hPr
ivat
eC
onsu
mpt
ion
10.
919
1.00
31.
086
1.25
2??
1.15
70.
929
1.05
41.
111
1.23
9?1.
161
1.27
8??
1.03
61.
273??
1.17
6?1.
149?
1.17
2??
0.95
21.
271?
1.14
81.
076
20.
983
1.06
91.
178
1.20
51.
111
0.96
71.
050
1.24
31.
241
1.13
71.
228
1.00
11.
219??
1.21
61.
178
1.13
50.
990
1.18
8?1.
200
1.17
4
30.
916
1.10
51.
072
1.19
1?1.
097
0.87
31.
065
1.12
41.
211?
1.12
71.
126
1.13
11.
155?
1.16
5?1.
191?
1.02
21.
108
1.13
5?1.
142
1.17
5
40.
770
1.24
20.
914
1.15
0?1.
078
0.77
71.
235
0.94
51.
124??
1.09
31.
017
1.11
61.
104
1.15
71.
124
0.92
51.
028
1.05
11.
096
1.06
7
hD
urab
leC
onsu
mpt
ion
11.
003
0.97
90.
966
0.97
90.
959
1.00
00.
988
0.98
20.
945
0.94
71.
012
1.02
21.
036
1.00
31.
017
0.99
71.
016
1.02
40.
982
1.00
1
21.
009
1.01
71.
017
0.94
70.
973
1.01
61.
015
1.02
50.
946
0.97
01.
048
0.99
81.
050
1.02
00.
996
1.03
80.
998
1.04
61.
017
0.99
2
31.
019
1.01
11.
018
1.01
60.
995
1.02
01.
005
1.02
71.
011
0.98
71.
058
0.99
61.
067
1.01
80.
996
1.04
71.
002
1.07
3?1.
014
0.99
7
41.
013
1.00
01.
019
1.06
81.
020
1.01
71.
003
1.02
01.
069
1.02
01.
071?
0.99
21.
055
1.04
81.
015
1.05
80.
993
1.05
91.
046
1.01
1
hN
on-d
urab
leC
onsu
mpt
ion
10.
921
1.05
31.
059
1.12
8?1.
135
0.92
31.
069
1.08
31.
104
1.11
2?1.
300
0.88
41.
178?
1.03
11.
006
1.20
50.
809
1.21
9??
1.01
50.
956
20.
957
1.17
01.
088
1.12
21.
118
0.93
41.
161
1.13
61.
128
1.12
51.
293
0.98
31.
179??
1.09
91.
095
1.22
70.
932
1.15
4?1.
056
1.06
1
30.
849
1.28
80.
964
1.03
31.
081
0.82
5?1.
293
0.99
71.
017
1.08
31.
149
1.09
41.
049
1.02
11.
038
1.07
30.
964
1.04
00.
968
0.98
7
40.
697??
1.47
5???
0.80
80.
967
1.04
40.
694??
1.48
2???
0.82
30.
934
1.04
30.
932
1.20
1??
0.97
51.
014
1.03
90.
881
1.09
00.
936
0.94
00.
993
hG
ross
Fixe
dC
apit
alFo
rmat
ion
11.
026
0.97
91.
038
1.02
30.
998
1.02
81.
009
1.04
11.
029
1.01
21.
064
1.03
90.
956
0.99
41.
022
1.05
21.
008
0.95
81.
010
1.02
4
20.
998
0.96
71.
012
0.99
90.
998
1.00
70.
991
1.03
31.
011
1.00
91.
064
1.02
90.
961
1.01
21.
040
1.04
01.
013
0.95
2?1.
027
1.04
7
31.
034
1.00
81.
000
0.97
31.
026
1.03
01.
002
1.00
80.
967
1.01
21.
021
1.04
31.
018
1.07
1??
1.08
6???
1.01
71.
056
1.01
81.
088?
1.11
7??
41.
011
0.98
91.
008
0.96
81.
013
1.01
50.
985
1.01
00.
962
1.00
21.
051
1.02
81.
001
1.03
31.
057??
1.03
51.
048
0.99
91.
054
1.09
0??
hC
onst
ruct
ion
and
Wor
ks
11.
025
1.01
41.
015
1.00
81.
001
1.02
21.
014
1.02
01.
022
1.00
60.
917
1.00
51.
004
1.03
21.
010
0.91
1?0.
986
1.00
21.
021
0.99
3
21.
026
0.99
30.
987
0.99
30.
982
1.01
70.
993
0.99
01.
007
0.98
40.
930
0.99
21.
011
1.03
61.
003
0.93
10.
993
1.00
21.
014
0.98
4
31.
042
0.99
40.
992
0.96
60.
980
1.03
40.
993
0.98
40.
981
0.98
10.
948
0.99
21.
019
1.04
71.
001
0.95
31.
001
1.00
51.
019
0.98
9
41.
040
0.99
90.
994
0.94
20.
985
1.03
40.
997
0.98
40.
960
0.98
90.
984
0.98
81.
018
1.04
70.
996
0.98
50.
997
0.99
61.
013
0.98
5
hM
achi
nery
and
Equi
pmen
t
10.
999
0.97
81.
049
1.00
31.
001
1.00
21.
004
1.04
30.
997
1.00
71.
033
1.00
20.
962
0.99
01.
002
1.02
30.
981
0.96
31.
000
1.00
4
20.
978
0.96
31.
046
1.03
41.
010
0.98
20.
987
1.05
81.
028
1.01
91.
053
1.00
20.
983
0.99
91.
012
1.03
60.
979
0.97
71.
024
1.01
4
30.
986
0.94
01.
056
1.07
81.
018
0.99
00.
939
1.07
61.
046
1.00
50.
986
0.99
91.
070??
1.04
31.
037
0.98
61.
000
1.10
6??
1.07
71.
063
40.
976
0.92
51.
039
1.06
31.
015
0.98
20.
930
1.04
31.
047
1.00
51.
008
1.02
11.
041
1.02
61.
029
0.99
91.
023
1.07
01.
064
1.05
3
(?)R
elat
ive
RM
SFE2 h
(see
equa
tion
3).O
rang
e-sh
aded
cells
=Fig
ures
belo
w1.
Gre
en-s
hade
dce
lls=
Figu
res
belo
w1
and
stat
isti
cally
sign
ifica
nt.
Har
vey,
Leyb
ourn
e,an
dN
ewbo
ld(1
997)
test
’sp-
valu
e:(???
)p<
1%,(??
)p<
5%,a
nd(?
)p<
10%
.Sam
ple:
2015
.I-20
19.II
I(19
obse
rvat
ions
).So
urce
:Aut
hors
’cal
cula
tion
s.
33
Documentos de Trabajo
Banco Central de Chile
NÚMEROS ANTERIORES
La serie de Documentos de Trabajo en versión PDF
puede obtenerse gratis en la dirección electrónica:
www.bcentral.cl/esp/estpub/estudios/dtbc.
Existe la posibilidad de solicitar una copia impresa
con un costo de Ch$500 si es dentro de Chile y
US$12 si es fuera de Chile. Las solicitudes se
pueden hacer por fax: +56 2 26702231 o a través del
correo electrónico: [email protected].
Working Papers
Central Bank of Chile
PAST ISSUES
Working Papers in PDF format can be downloaded
free of charge from:
www.bcentral.cl/eng/stdpub/studies/workingpaper.
Printed versions can be ordered individually for
US$12 per copy (for order inside Chile the charge
is Ch$500.) Orders can be placed by fax: +56 2
26702231 or by email: [email protected].
DTBC – 887
Railroads, specialization, and population growth in small open economies: Evidence
from the First Globalization
Andrés Forero, Francisco A. Gallego, Felipe González, Matías Tapia
DTBC – 886
High Dimensional Quantile Factor Analysis
Andrés Sagner
DTBC – 885
Heterogeneous Paths of Industrialization
Federico Huneeus, Richard Rogerson
DTBC – 884
Does the Commodity Super Cycle Matter?
Andrés Fernández, Stephanie Schmitt-Grohé, Martín Uribe
DTBC – 883
Twitter-Based Economic Policy Uncertainty Index for Chile
Andrés Sagner, Juan Sebastián Becerra
DTBC – 882
Corporate-Sector Functional Currency: An International Comparison
Jorge Fernández, Fernando Pino, Francisco Vásquez
DTBC – 881
Back testing fan charts of activity and inflation: the Chilean case
Jorge Fornero, Andrés Gatty
DTBC – 880
Financing Firms in Hibernation during the COVID-19 Pandemic
Tatiana Didier, Federico Huneeus, Mauricio Larrain, Sergio L. Schmukler
DTBC – 879
Choice Aversion in Directed Networks
Jorge Lorca, Emerson Melo
DTBC – 878
Big G
Lydia Cox, Gernot Muller, Ernesto Pasten, Raphael Schoenle, Michael Weber
DTBC – 877
Sticky Capital Controls
Miguel Acosta-Henao, Laura Alfaro, Andrés Fernández
DTBC – 876
Measuring the perceived value of an MBA degree
Carlos Madeira
DTBC – 875
The Real Effects of Monetary Shocks: Evidence from Micro Pricing Moments
Gee Hee, Matthew Klepacz, Ernesto Pasten, Raphael Schoenle
DTBC – 874
Measuring Systemic Risk: A Quantile Factor Analysis
Andrés Sagner
DTBC – 873
The impact of information laws on consumer credit access: evidence from Chile
Carlos Madeira
DOCUMENTOS DE TRABAJO • Septiembre 2020