pierpaolo parrotta and dario pozzoli - connecting repositories · working paper 09-11 pierpaolo...
TRANSCRIPT
WORKING PAPER 09-11
Pierpaolo Parrotta and Dario Pozzoli
The Effect of Learning by Hiring on Productivity
Department of Economics ISBN 9788778823946 (print)
ISBN 9788778823953 (online)
The Effect of Learning by Hiring on Productivity∗
Pierpaolo Parrotta †
Aarhus School of Business, Aarhus University
Dario Pozzoli‡
Aarhus School of Business, Aarhus University
November 3, 2009
Abstract
This work studies the phenomenon of inter-firm labor mobility as potentialchannel of knowledge transfer. Using data from the Danish employer-employeeregister, covering the period 1995-2005, it investigates how the knowledge embed-ded into recruited workers, coming from other firms, contributes to the processof knowledge diffusion and enhances firm productivity. Specifically, estimatingboth parametric and semi-parametric production functions (Olley and Pakes,1996; Levinsohn and Petrin, 2003), the impact of recruited technicians and highlyeducated workers on total factor productivity at the firm level is found to be sig-nificantly positive. A matching analysis, which allows for continuous treatmenteffect evaluation (Hirano and Imbens, 2004), corroborates this finding.
JEL Classification: C23, J33, J38, J51Keywords: Labor mobility, Total factor productivity, Knowledge transfer
∗We thank Michael Rosholm, Hans Kongsted, Tor Eriksson, Lars Geerdsen, Davide Sala, SanneHiller, Frederic Warzynski and Valerie Smeets for helpful suggestions. We also thank the CEBR forthe provision of data on Danish patent application ascribed at EPO. The usual disclaimer applies.†Corresponding Author. Aarhus School of Business, Aarhus University, Department of Economics,
Hermodsvej 22, DK, 8230 Aabyhoj, Denmark. E-mail: [email protected]‡Aarhus School of Business, Aarhus University, Department of Economics, Hermodsvej 22, DK,
8230 Aabyhoj, Denmark. E-mail: [email protected]
1
1 Introduction
Knowledge can be defined as the fundamental resource employed in all valuable hu-
man activities. Its identification as economic good has requested a long time and
enduring efforts (Antonelli, 1995). A group of models belonging to the so-called New
Growth Theory represents the attempt to understand and describe the role of knowl-
edge in driving productivity and thereby fostering economic growth. Specifically, sev-
eral macroeconomists (Romer 1990, Grossman and Helpman 1991, Aghion and Howitt
1992, among the pioneers) have identified knowledge spillovers as the main driver of
sustained economic growth.
A potential channel of knowledge transfer is the inter-firm labor mobility. Most of
valuable knowledge, characterized by a tacit and complex nature, is embedded in highly
educated or skilled workers. Thus, moving from one firm to another, such individuals
carry this knowledge and apply it to new contexts. The acquired knowledge hardly can
be transmitted in a different way since it tends to be stably localized within firms. Given
the stickiness in routines and procedures, its transfer cannot be effectively prevented by
patents or other forms of Intellectual Property Rights, which can only protect forms of
codified or at least articulate knowledge (Petit and Tolwinski 1996, Gilson 1998). This
process of knowledge transmission via inter-firm labor mobility is known as learning
by hiring (LbH, henceforth).
In literature very little evidence exists about the effect of such process on firm
performance. Previous works found a positive impact of specific categories of recruited
workers on the firms patenting activity. However, the distribution of patenting firms is
strongly concentrated and the fruits of knowledge transfers may or may not be related
to the patenting activity. Instead, they surely may affect the production process.
Thus, it seems opportune to test the effects of LbH on firms productivity, which is
surely a broader measure of performance. This variable allows to take into account a
2
huge number of firms and therefore to work on a quite representative sample of the
production and service sectors.
This paper opens to the modeling of the LbH and investigates how the composition
of newly recruited workers, coming from other enterprises, affects the productivity in
the arrival firm. Specifically, the analysis is focused on the role of who experienced
a wage increase after moving (from the donor firm) and was characterized by a ter-
tiary (Bachelor, Master or Post-Graduate Degree) or vocational/technical education.
The higher wage offered to the newly employed may represent a signal of precise will-
ingness in her enrolment shown by the firm. That can be interpreted as a necessary
condition for the identification of the incoming knowledge carriers, who do not sim-
ply replace retired or fired workers. The sufficiency is instead fulfilled by the above
mentioned employees’ highest educational attainments. However, the plausibility of
a given knowledge transfer does not inform also on the nature of its receipt. That
may depend on two relevant features: intensity and proximity. The latter refers to the
technological distance between the donor and recipient firm (Jaffe, 1986; Adams, 1990;
Inkmann and Pohlmeier, 1995; Cincera, 2005). It is the distance in a metric space
generated by vectors, whose elements are firm characteristics. Instead, the intensity is
a concave function of the number of incoming workers from a given firm.
In the empirical analysis both parametric and semi-parametric production functions
(Olley and Pakes, 1996; Levinsohn and Petrin, 2003) are estimated. Measures of Total
Factor Productivity (TFP henceforth) are successively obtained as residuals and then
regressed on a variable representing the knowledge inflow. The effects are positive
and significant. The contribution of the LbH to firms productivity is also confirmed by
implementing a matching analysis. Specifically, allowing for continuous treatment effect
evaluation (Hirano and Imbens, 2004), the treatment effects show a positive impact on
productivity. Overall the results are consistent with New Growth theoretical models,
which place a strong emphasis on knowledge diffusion, and may partially explain the
3
increase in the TFP recorded in Denmark during the last decades. In particular, they
suggest improvements in modeling the process of knowledge transmission at micro
level, stimulating further analysis on the additional effects associated with the inter-
firm labor mobility.
The structure of the paper is as follows: section 2 briefly reviews the relevant
literature, section 3 describes the data, section 4 provides details on the empirical
strategy, section 5 illustrates main results, section 6 shows the sensitivity analysis, and
section 7 concludes.
2 Literature background
The process known as LbH refers to the valuable knowledge transferred by inter-firm
labor mobility. It is defined as the acquisition of knowledge from other firms through
the hiring of experts (Song, Almeida and Wu, 2003). In accordance with this defini-
tion, benefits of LbH may not be confused with either the usual high labor productivity
characterizing the newly employed experts, or potential externalities associated with
scale effects. In such context, it is worth emphasizing that inflows of knowledge, car-
ried by some categories of workers, are typically forms of developed but not codified
knowledge. As a consequence, what carried by mobile qualified workers cannot spill
over freely. Hence, the impact of LbH can be precisely quantified only if a donor and a
recipient firm are identified. This clearly implies that in absence of a matched employer-
employee register is not possible to estimate the contribution of this knowledge transfer
to any kind of firm performance indicator. However, even after the availability of the
cited registers, the use of inter-firm labor mobility as learning mechanism has not
received particular attention.
Although their study is not conducted at firm level, Almeida and Kogut (1999) find
4
that the mobility of engineers holding major patents affects the intra- and inter-regional
pattern of patent citations, which is considered as a proxy of knowledge flow. Mapping
the US semiconductor plant clusters through the use of county level establishment
and employee data, they illustrate how the mobility of engineers partially explain
innovations occurred among regional clusters.
A step forward is moved by Rosenkopf and Almeida (2003). Tracking both inter-
firm mobility of engineers and patent citations for the second wave of entrants in the
US semiconductor industry, their analysis concludes that the effectiveness of labor mo-
bility in terms of knowledge flows increases with the degree of technological distance
between firms. These findings are confirmed and even reinforced by Song, Almeida and
Wu (2003). In this work they investigate the conditions under which labor mobility is
more likely to facilitate knowledge transfers. Results seem to support the hypotheses
that LbH is more effective (in terms of contribution to patent production) in case (i)
the recipient firm is less path dependent;1 (ii) the mobile engineers carry knowledge
distant from that characterizing the recipient firm; (iii) the knowledge carriers work
in noncore technological areas in the arrival firm. Thus, inter-firm labor mobility may
mitigate the difficulties of learning from firms characterized by research in distant tech-
nological areas and increase the possibility to benefit from external knowledge.
A different perspective is instead investigated by Kim and Marschke (2005). Using
a sample of US firms, they analyze how the risk of a key employee’s departure reduces
the firm’s R&D expenditure and/or increases its patenting propensity. Their findings
are consistent with the statement that firms use patents to minimize the negative
effect associated with the departure of an engineer or scientist. Moreover, that may
help explaining what determines different propensity to patent among firms. Further
1The path dependence is measured as number of self-cited patents. Self-citing occurs when a patentfiled by a firm cites another patent from the same firm (Song, Almeida and Wu, 2003).
5
evidence that R&D workers transmit valuable knowledge is provided by Maliranta,
Mohnen and Rouvinen (2009). Using Finnish employer-employee data, the authors
find that firms hiring workers previously engaged in R&D activities may increase their
productivity and profitability if they do not perform any R&D activity. This finding is
interpreted as evidence that mobile R&D workers may transmit knowledge, which can
be readily copied and implemented without particular research efforts.
A crucial study for the main purpose of this paper is Kaiser, Kongsted and Rnde
(2008). They constructed a dataset in which patent applications sent by Danish firms
to European Patent Office (EPO, henceforth) are matched to the employer-employee
register. Their research question is how labor mobility affects innovation in Danish
firms, and more in detail how the composition and past patenting experience of labor
inflows influence the firm-level patenting activity. They split the firm workforce into
”R&D workers” and ”non-R&D workers”. Whereas the former class is composed of
employees with Bachelor’s or Master’s degree in natural in technical fields, the latter
one identifies individuals with the same level of education but in humanities. The
results of their analysis support the idea that mobile R&D workers contribute more
to the firm patenting activity than immobile R&D employees. This effect is stronger
in case the R&D worker has been previously hired by a patenting firm. However,
they find weak evidence that R&D employees carry a larger amount of knowledge than
employees with other qualifications.
Although Kaiser, Kongsted and Rnde (2008) assess more formally the quantitative
effects of labor mobility on innovations, their paper does not differ from the other men-
tioned studies for the variable indicating the firm performance. This common feature
is the focus on the patenting activity that is a proxy for innovation efforts. There-
fore, they constrain the amount of knowledge transferred only to patent applications
or grants. These may represent a measure of codified, rather than tacit or less articu-
lated, knowledge created by a firm: fruits of knowledge transfers could or could not be
6
related to patents. Moreover, since the patenting activity is particularly concentrated
among firms (this holds even strongly in Denmark) the number of firms observed in
the sample may not be representative of the production activities in which a relevant
part of valuable knowledge sticks.
3 Data
3.1 Data Sources
The dataset used in the empirical analysis was constructed by merging information
from three different main sources. The first data source is the ”Integrated Database
for Labor Market Research” (IDA henceforth) provided by Denmark Statistics. IDA
is a longitudinal employer-employee register containing valuable information (age, de-
mographic characteristics, education, labor market experience, work experience and
earnings) on each individual employed in the recorded population of Danish firms dur-
ing the period 1980-2005. Apart from deaths and permanent migration, there is no
attrition in the dataset. The labor market status of each person is recorded at the 30th
of November each year. The retrieved information has been aggregated at firm level
and consequentially merged to variables like enterprises’ location (County), size and
related industry.2
The second data source refers to firms business accounts (REGNSKAB henceforth),
which has been also provided and compiled by Denmark Statistics. It covers the con-
struction industry from 1994, manufacturing from 1995, wholesale trade from 1998 and
the remaining part of the service industry from 1999 onwards. Data in REGNSKAB is
a number of aggregations of yearly financial items, which are crucial for the estimation
of the production function. In particular, it is possible to retrieve information on sales,
2In our empirical analysis we exclude from the analysis the following sectors: i) agriculture, fishingand quarrying; ii) electricity, gas and water supply and iii) public services.
7
intermediate goods or materials, fixed assets, and profits. Although statistics in REG-
NSKAB have been gathered in several ways, part of them refers to selected firms for
direct surveying: all firms with more than 50 employees or profits higher than a given
threshold. The other firms are recorded in accordance with a stratified sample strat-
egy. The surveyed firms can choose whether submit their annual accounts and other
specifications or fill out a questionnaire. In order to facilitate responding, questions
are formulated in the same way as required in the Danish annual accounts legislation.
Finally, further information is gathered from a third dataset on patent applications
and grants ascribed to Danish firms at the EPO in the period 1978-2003. The access
to such data has been made possible thank to the Centre for Economic and Business
Research (CEBR).3 A total of 12,109 patent applications have been recorded. Being
available the unique assigned identifier, 2,822 Danish non-person patent applicants have
been identified.4 That allows computing measures of technological proximity partially
based on patent applications.
3.2 Variables
This section describes the variables used in the empirical analysis. Since the purpose of
this paper is to provide evidence concerning the impact of LbH on firms productivity,
the attention is devoted almost completely to the explanation of the measures indicating
inter-firm knowledge transfers via labor mobility.
Combining the needed information from the three data sources previously described,
the final dataset allows for the mapping of all mobile workers aged 18 to 60. For
the identification of knowledge carriers, a set of necessary and sufficient conditions is
imposed. Whereas the sufficiency is provided by the attainment of either a tertiary (at
3An independent research centre affiliated with the Copenhagen Business School (CBS).4More details concerning the construction and composition of the dataset can be found in Kaiser
et al. (2005).
8
least Bachelor’s degree) or a vocational/technical education, the necessary conditions
are the following ones: i) the mobile worker has to experience a real5 wage increase
of 5 per cent after moving; ii) the real annual wage of the current job should be
higher than the average of real wages received in the last three or two years; iii) the
incomers’ wage should be greater than the average wage recorded in the recipient
firm; iv) the sending firm is not downscaling the labor force6; v) the eventual period
of unemployment preceding the start of the new job should be less than 3 months.
The necessary conditions can be interpreted as a signal of precise and intentional
recruitment strategy implemented by the recipient firm. In fact, it is here assumed that
every recipient firm, being aware of the expected benefit deriving from the knowledge
embodied into some educated and skilled workers, is willing to pay an opportune wage
premium.
Three measures of inter-firm knowledge transfers have been defined.
The first measure is simply constructed as sum of all incoming workers kcj fulfilling
all previous necessary conditions from the donor firm d to the recipient r:
LbHr =D∑
d=1
J∑j=1
kcjdr
Though quite similar to the first measure, the second one weights the number of
incoming workers from the same donor firm using a concave function. The concavity is
ensured by the inverse of the total of qualified workers KCd (highly educated employees
and technicians):
5The annual wage is deflated using the price deflator for the year 2000.6The yearly reduction in the total labor force is imposed to be smaller than 1 per cent.
9
LbH − concaver =D∑
d=1
(J∑
j=1
kcjdr
)1/KCd
It obviously implies that LbH − concaver increases less than proportionally with
respect to the total number of qualified workers leaving a given donor firm and em-
ployed in the same recipient firm. The first and second measure coincide just in case
single workers (from different donor firms) move towards a given arrival firm. This
means that the concave version adjusts for the intensity of each inter-firm knowledge
inflow: two qualified workers may transfer together from the same firm less than the
double amount of knowledge that a single one could transfer. Hence, a large amount
of knowledge embodied in procedures or production processes is brought by the first
mover. The second measure attributes more public good features to the knowledge
inflow, even though it is still characterized by partial-excludability7 and low degree of
appropriability.
The third and most complete measure introduces a further element represented by
the proximity index ψ:
LbH − proxr =D∑
d=1
ψ(d, r)
(J∑
j=1
kcjdr
)1/KCd
That allows to correct the intensity of inter-firm labor mobility for the technologi-
cal distance between the donor and recipient firm (Jaffe 1986, Adams 1990, Inkmann
and Pohlmeier 1995, Cincera 2005). The assumption behind this index is that the
7Non-excludability means that once a good has been produced, it is not possible to prevent otherpeople from gaining access to it (more realistically, it is costly for the provider to exclude unauthorizedusers).
10
technology developed by a firm can affect the productivity of other firms, even though
no transactions of intermediate or capital goods occur. Thus, the level of exploitabil-
ity associated with the labor mobility could depend on the degree of similarity in the
technological knowledge characterizing the two firms. In order to measure the distance
between firms technological capabilities, a technological space needs to be defined. In
the present work, the generator vector f is composed of elements reflecting the follow-
ing firm characteristics: shares of highly educated workers and technicians, stock of
patent applications in different technological areas,8 related industry and firm size cat-
egories. The technological distance is here computed using the uncentered correlation
suggested by Jaffe (1986):
ψdr =fdf
′r[(
fdf′d
)(frf
′r)]1/2
If the donor and recipient firm coincide perfectly in the generated technological
space, then ψ = 1. Conversely, If they do not overlap at all, the weight takes on the
value 0. However, it is worth underlining that the firm measures can be only used for
patenting firms. In fact, the inclusion of the non-patenting firms biases the technolog-
ical space: a mass of firms would present a proximity determined exclusively by the
non-patenting behaviour.
Besides the three described measures, the variables used in the empirical analysis
are: valued added, materials capital stock, labor force, dummies for counties9, indus-
tries and years. Using the price deflator for the year 2000, the monetary values have
8The technological areas are: Chemicals, Consumer goods, Machinery, Process engineering andOther.
9The counties are: Copenhagen, Freriksborg, Roskilde, West Zealand, Storstroms, Bornholms,Funen, South Jutland, Ribe, Vejle, Ringkobing, Aarhus, Viborg, North Jutland.
11
been converted in real terms.
3.3 Descriptive statistics
Once depicted data composition and defined variables, the exposition continues showing
some descriptive statistics. Table 1 reports number of observations, median, mean
values and standard deviations of the main variables for the whole sample and by firm
size. This is classified into three categories (size1, size2 and size3), which are referred
respectively to firms with less than 50, between 50 and 99 and more than 100 employees.
A huge number of observations is recorder over time and more than its 95 per cent is
represented only by the category size1.
This percentage is quite consistent with the Danish industrial structure (private
sector), which is largely dominated by small sized enterprises.10 Although not all indus-
tries have been recorded from the first years of the sample period, percentages reported
among industries reflects partially aggregate data for the same time span. Whereas
observations for construction and manufacturing industry are over-represented, the op-
posite is shown by those for financial and business sector. However, that is entirely
explained by the unbalanced structure: construction and manufacturing have been the
only industries recorded in the first years, whilst financial and business just in the last
ones.
Looking at values of accounting (reported in thousands of DKK) and workforce
variables, a considerable distance between median and mean values can be easily rec-
ognized. That points out a certain degree of skewness (and polarization) among firms
in the sample, confirmed also by the notable dimension of the standard deviations.
10The structure of the Danish firm population is mainly composed of small and medium sized com-panies. Enterprises with less than 50 employees account for 97 per cent of the total number of firmsand represent 42 per cent of the total employment in manufacturing and services. Thus, it seems rea-sonable to distinguish between SME and larger enterprises using a limit of 100 employees. Comparedto the other Nordic countries, Denmark presents a fewer micro firms (between 1-9 employees): theserepresent 87 per cent of firms, against 92 and 94 per cent respectively recorded in Finland and Sweden(OECD, 2005).
12
Descriptive statistics change sensibly among the three definitions of inter-firm labor
mobility. The count measure presents the highest mean values, instead the measure
including the proximity index shows the lowest ones. The three measures increase their
dimension with respect to the firms size.
4 Empirical Strategy
4.1 Productivity estimation
The firm productivity is obtained from a Cobb-Douglas production function con-
taining the real value added (Y), labor (L), capital (K) and a number of other controls
affecting productivity, such as firm specific characteristics of employees, foreign own-
ership, year, size and regional dummies (Z). Since input characteristics differ across
industries, production function parameters are estimated for each 2-digit sector j sep-
arately.11 Therefore, our reported results use the following specification:
ln(Yijt) = cons+ aln(Lijt) + bln(Kijt) + c(Zijt) + uit, (1)
where the error term uit consists of an unobserved time-invariant firm effect νi and
an idiosyncratic component εit.12 The panel data setting allows for the computation
of within estimators, which may remove the bias related to the omission of firm fixed
effects. However, FE within estimators are biased or even inconsistent in case levels
of inputs and output are chosen simultaneously or explanatory variables are affected
by measurement errors (Griliches and Hausman 1986). Thus, it seems opportune to
11Value added and capital stock have been deflated by using GDP deflators drawn from World BankIndicators (base year 2000).
12The value added has been calculated as suggested by Denmark Statistics. We use the annualmeasure of the capital stock reported in the REGNSKAB.
13
compare FE within estimation with two alternative semi-parametric approaches: the
Olley-Pakes (1996) and the Levinsohn-Petrin (2003a) estimation methods. The Olley-
Pakes (OP henceforth) estimation may correct not only for simultaneity but also for
the selection bias resulting from the relationship between productivity shocks and the
probability of exit from the market. Specifically, it is assumed that the error term, uit,
is the sum of two shocks:
uit = Ωit + ηit, (2)
where Ωit13 is the productivity shock observed by the firm but not by the econome-
trician, and ηit is an unexpected productivity shock, which is unobserved by both. The
assumption on uit implies that inputs are correlated with the realization of productiv-
ity shocks (simultaneity issue). In addition, if profitability is positively related to Kit,
then a firm with a higher capital stock will expect larger profits at current productivity
levels; else lower profits will be associated with the exit from the market (Yasar et al.
2008), especially for the smaller firms. In the OP approach, incumbent firms decide, at
the beginning of each period, whether continue participating in the market and then
choose levels of investment and inputs employed. Hence, the OP method first requires
the modeling of firm’s investment decision:
Iit = I(Ωit, Kit). (3)
Provided that function I(.) is positive, strictly increasing and there exists a positive
correlation between its arguments, it is possible to obtain the inverse function for the
observed shock variable as13Ωit is assumed to follow a first-order Markov process.
14
Ωit = I−1(Iit, Kit) = h(Iit, Kit). (4)
This function is useful to deal with the simultaneity issue. In fact, substituting (2)
and (4) into (1), it yields
ln(Yit) = aln(Lit) + cln(Zit) + φ(ln(Iit), ln(Kit)) + ηit (5)
φ(ln(Iit), ln(Kit)) = cons+ dln(Kit) + h(Iit, Kit), (6)
where cons and d are parameters and function h(.) is approximated by a second-
order polynomial series in capital and investment. The identification of d requires the
estimation of survival probabilities, which will then allow us to control for selection
bias. Specifically, firm i decides to remain active or exit the market if its productiv-
ity is respectively greater or less than a given threshold, which depends on past and
current levels of capital stock. Following Yasar et al. (2008), the survival probability
has been estimated by fitting a probit model on Iit and Kit, their squares and cross
products. These predicted probabilities are here called Pit. Finally, nonlinear least
square estimates will be computed for parameters in the following equation
ln(Yit) = aln(Lit) +
c(Zit) + bln(Kit) + g(φt−1 − bt−1ln(Ki,t−1), Pit) + ηit, (7)
where the unknown function g(.) is approximated by a second-order polynomial
15
in (φt−1 − dt−1ln(Ki,t−1)) and Pit. A major drawback of the OP estimation method is
that investment proxy may not smoothly respond to productivity shocks, because of
adjustment costs, violating thus the condition for consistency. Moreover, the invest-
ment proxy is only valid for plants reporting non-zero investment level, inducing de
facto a truncation bias.
To overcome these limits, Levinsohn and Petrin (2003a) (LP henceforth) suggest to
use intermediates instead of investment for the estimation of firms production function.
Since many intermediates are almost always non-zero, LP approach circumvents the
above mentioned data truncation problem. Furthermore, adjustments in intermediate
inputs may respond to the entire productivity term more smoothly than OP’s invest-
ment proxy: they are typically less costly. LP model is more complex to program than
OP procedure. However user-friendly Stata programs are available for both estimation
methods. Using the estimates of production function parameters, the firm i ’s TFP, at
time t in industry j, is defined as
lnTFPijt = ln(Yit)− ajln(Lit)− bjln(Kit) (8)
Next to the computation of TFP values, the relationship between these and alter-
native measures of LbH can be estimated in the following equation:
ln(TFPijt) = α + β1ln(LbHt) + β2ln(LbHt−1) + γz(Zt) + ξit,
where β1 and β2 are respectively the contemporaneous and lagged LbH effects asso-
ciated with the inflows of qualified workers and Zt represents firm specific characteristics
16
of employees,14 whether the firm is foreign owned and a full set of size, time, regional
and industry controls.
4.2 Treatment effects based on the generalized propensity score
In this section, it is described an alternative econometric methodology to evaluate
the effect of LbH on firm productivity. In particular, following the latest developments
in the treatment evaluation literature, it is here compared the TFP of firms exposed
to LbH with ”matched” less exposed firms.15 This approach interprets the inflow of
qualified workers as treatment and evaluates its effect on firm TFP.
As above cited, TFPijt is the outcome variable and Xi,t−1 is the vector of pre-
treatment characteristics. As Hirano and Imbens (2004) suggest, we define a set of
potential treatment values Dit and the unit-level dose-response function TFPijt(d)d∈D.
Thus, it is assumed that (a) TFPijt(d)d∈D, Dit, and Xi are defined on a common
probability space, (b) Dit is continuously distributed. To simplify the notation, the
subscripts i, j and t are erased. The propensity to receive the treatment and the
generalized propensity score (GPS, henceforth) are defined respectively as r(d, x) =
fD|X(d|x) and R = r(D,X). Furthermore, the GPS is required to fulfill the following
balancing property:
X⊥1(D = d)|r(d, x), (9)
where 1(.) is the indicator function. According to Hirano and Imbens (2004)
14Considering the possibility that firms may benefit from a variation in the skill composition of thelabor force, we include the shares of highly skilled workers and technicians among the firm specificcharacteristics of employees.
15The approach allows to estimate the effects between units at different treatment levels, by com-paring the values of the estimated potential outcome for different levels but not an estimation of theeffects between treated and non-treated units.
17
such balancing property, in combination with a suitable CIA assumption, implies that
assignment to treatment is unconfounded, given the GPS.
The implementation of the GPS matching method consists of three steps (Bia and
Mattei, 2008). In the first step, the score R is estimated and the treatment (or a
monotone transformation of it) is required to respect a normal distribution
g(D)|X ≈ N [h(γ,X), σ2], (10)
where g(D) is a suitable transformation of the treatment variable and h(γ,X) is a
function of covariates with linear and higher-order terms, which depends on a vector
of parameters, γ. 16
In the second step, the conditional expectation of the outcome, TFP , given D and
R, is modelled as:
E(TFP |D,R) = α0 + α1D + α2D2 + α3D
3 + α4R + α5R2 + α6R
3 + α7DR. (11)
whose parameters are estimated by applying ordinary least squares. As Hirano
and Imbens (2004) emphasize, there is no direct meaning in the estimated coefficients
of the selected model, except for testing whether the covariates introduce any bias.
Given the parametric model used to compute the GPS, it is possible to prove the
root-N consistency and asymptotic normality of the estimator. Standard errors and
confidence intervals are here obtained by employing a bootstrapping procedure.
16The potential inclusion of the higher-order terms is aimed at the satisfaction of the balancingproperty.
18
The last step consists of averaging the estimated regression function over the score
function evaluated at the desired level of treatment (Bia and Mattei, 2008). Specifically,
in order to obtain an estimate of the entire dose-response function, we estimate the
potential average outcome for each level of the treatment of interest as
E ˆTFP (d) =1
N
N∑i=1
βd, d(d,Xi) =1
N
N∑i=1
ϕ−1[ψd, d(d,Xi); α], (12)
where α is the vector of the parameters estimated at the second stage.
5 Results
This section illustrates and discusses results obtained by the implemented empirical
analysis. Firstly, the attention is devoted to the estimation of the TFP level and how
such estimate is associated with the LbH measure. Secondly, the analysis moves to
the investigation of potential causal effect of LbH on firm productivity. As outlined in
section 4.1, three different approaches (FE, OP and LP) have been followed in order to
estimate productivity at firm level. As the OP approach requires positive investment
values to estimate the production function coefficients, it has been possible to estimate
different production technologies at one-digit industry codes only. The 2-digit sector
specific elasticities for capital and labor stock, estimated by using FE and LP methods
are reported in Table 2. Table 3 shows results from the OP approach. Whereas the OP
and LP estimations lead to more similar results, the labor (capital) coefficients seem to
be in many cases underestimated (overestimated) with FE. Once estimated production
function parameters, the TFP is therefore computed like a residual. The log of TFP is
then regressed (by using OLS) on the LbH measure and its lag, including a large set of
19
controls.17 In these regressions, residuals are clustered at firm level. The introduction
of the LbH lag variable is reputed opportune to capture some dynamics of the knowl-
edge transfer effect: it may take some time to be effective. Parameters associated with
LbH show significant and positive coefficients. They differ among specification and es-
timation methods. This implies that the inclusion of intensity and proximity features
change the size of coefficients. Specifically, in the pooled regressions count measures
present the highest values whereas the opposite is true for the proximity measures.
These have been computed on a smaller sample, as shown in Table 4 it accounts of
1.2 per cent of the total number of observations, which includes all firms that sent at
least a patent application to the EPO. Except for few cases, the LbH lag influences
the firm TFP more than the current LbH value: it can be even twice times larger than
the contemporaneous effect. In order to facilitate the comparison between estimation
methods across industries, results from the FE and LP second stage estimation are also
presented at one-digit industry level18 in Table 5. It is here illustrated that LbH is a
common phenomenon and it does not occur only in specific industries, such as more
knowledge based or characterized by higher technological intensity. However, it varies
among industries: it does not appear particularly robust for Transports and Wholesale-
Retail Trade but significantly sizeable for Financial-Business and Construction sectors;
Manufacturing activities seem to be quite stably affected by LbH too. The reported
elasticities of firm TFP with respect to LbH measures are between 1 and 3 per cent
in the majority of cases. TFP estimates coming from LP method might be the most
precise ones. In fact, even though the OP semi-parametric approach deals with the
typical FE estimation problems due to simultaneity and endogeneity, it generates a
17The controls refer to the shares of managers, middle managers, females, highly educated workers,technicians, employees belonging to 4 age categories, whether the firm is foreign owned and a full setof year, size and county dummies.
18Results from second stage estimation at two-digit level can be provided by the authors to whomthese may interest.
20
data-truncation problem caused by zero-investment spells.19 The discussed regression
analysis sheds light on the plausibility of the positive impact of knowledge transfer via
inter-firm labor mobility on firm productivity. Contributions of capital and labor stock
and any other variable, included in the first stage estimation of production function, do
not affect the generated TFP values in the second stage. Furthermore, since other po-
tential factors are captured by the included controls, the contribution of LbH measures
to TFP may be correctly identified. It seems somehow hard to state that an increase
in TFP leads to recruit more workers with given characteristics instead of the opposite
causation direction. However, this interpretation is still possible to be claimed. //
In order to investigate the potential causal effect of LbH on firm productivity, a
matching analysis has been conducted. Specifically, the focus is here on the estimation
of the dose-response function and treatment effect behavior. The implementation of
such analysis is also particularly important since it may provide useful policy sugges-
tions: it allows us to evaluate the impact of knowledge transfer on productivity for
different level of LbH exposure. The outcome variable (log of TFP) is taken in lev-
els, the treatment is obviously the log of the LbH measure. The matching variables
include whether the firm is foreign owned, the shares of managers, middle managers,
females, highly educated workers, technicians and employees belonging to 4 age cate-
gories. All these characteristics have been considered the year before treatments occur.
We also add a set of year, size, industrial and regional dummies to control for com-
mon aggregated demand and supply shocks. Given the number of observations and
available counterfactuals, it has not been possible to estimate LbH effects for each
industry (even at 1-digit level), so such effects are evaluated in a pooled regression.
In addition, this lack of observations (Table 6) prevented us to estimate consistently
different treatment effects for each year. To assess the balancing property, we divide
19Implementing the OP estimation method, the number of observations is less than one third withrespect to the LP one.
21
the treatment variable into terciles and test whether the GPS adjusted mean differs
in one tercile compared to the others.20 Table 7 shows that the average dose-response
functions increase with respect to the level of treatment, confirming therefore the pos-
itive contribution of LbH to firm productivity. Doses of treatment produce significant
responses when the treatment reaches given levels in its distribution. Therefore, the
LbH effect is increasing but insignificant for quite low treatment levels, the effect of
recruiting qualified workers is significantly positive and stronger as doses of treatment
increases. The average treatment effects are here defined respectively as the variation
in the estimated response function due to a 10, 20, and 40 per cent change (delta) in
the treatment variable. As an instance, a 40 per cent increase in the treatment can
enhance at best the firm productivity by 0.12 percentage point. These findings seem to
be not completely in line with what found by implementing the (two stage) regression
approach. However, estimated LbH effects over different levels of treatment are not
easily comparable with the coefficients obtained in the previous analysis for a main
reason: the matching analysis here implemented does not include firms having zero
level of treatment in the control group. This obviously implies that the comparison in
the outcome variable takes place between similar treated firms only.
6 Sensitivity Analysis
In the definition of knowledge carriers several and strict conditions have been imposed
in order to capture as precisely as possible the real contribution of knowledge transfers
to firm productivity. However, being aware that the degree of strictness is a relative
20This is equivalent to testing that the conditional mean and treatment indicator are independent(CIA) where r(d;X) is evaluated at the median value of the treatment within the tercile d*. FollowingHirano and Imbens (2004), we test this hypothesis by blocking. For each tercile we define three blocksand compute the mean difference in X for observations (D=d) and (D). Then, we combine these threemean differences, weighted by the relative number of observations in each block, and compute theassociated t-statistic value.
22
concept, in this section we try to strengthen (refine) or relax (coarsen) the adopted
criteria and then check whether such changes influence the size and significance of
LbH effects. The focus is here on the TFP estimated from LP method since it does
not produce data truncation problem and potential simultaneity bias. In addition, we
decide to do not perform any robustness check for the GPS matching approach since
it is already subject to several properties to fulfill and the number of observations is
particularly low. Unambiguously, in all checks implemented results change size and
statistical significance as expected: they are shown in Table 9. In the first change,
the sample is restricted to all firms whose accounting variables have been not imputed
by Denmark Statistics. It is therefore evident that the estimates are affected by im-
putations very slightly even though the sample size shrinks. In the second one, the
distribution tails of LbH measures are not included in the regressions in order to pre-
vent any potential bias induced by the presence of outliers. Next, it is required a yearly
wage premium of 10 per cent for qualified incomers: this strengthens heavily the neces-
sary condition on the willingness of recruitment. Consecutively, the zero-downscaling
firm labor force condition is tested too. Finally, we exclude all knowledge carriers with
vocational education, degree in humanities and social sciences. In all refinements, the
coefficients related to LbH effects improve a little. Instead, from column 16 to 18 we
check whether the increase in knowledge carrier’s wage is the only driver of the firm
TFP enhancement: these results reject such suspicion. Although knowledge carriers
are typically highly productive, their contribution to firm TFP may not be imputed
to their labor productivity, which is often associated with the earned wage. From col-
umn 19 to 21, we exclude the knowledge carriers experiencing unemployment before
moving. In addition, we also investigate whether the exclusion of qualified workers
experiencing a transition from part-time to full-time regime affects the LbH impact. In
the last column, we weight more distance rather than proximity in firm characteristics,
simply replacing ψ(d, r) with (1-ψ(d, r)) in the third LbH index. It turns out that dis-
23
tance softens LbH effects, confirming our previous assumption and weakening partially
what claimed by Rosenkopf and Almeida (2003), and Song, Almeida and Wu (2003).
Moreover, it is also outlined that knowledge transfers are not determined exclusively
by workers having tertiary education in natural or applied sciences (as assumed in the
majority of papers in this literature), but also by those with degree in other fields or
by technicians.
7 Discussion and Conclusions
This paper investigates the consequences of recruiting qualified personnel on firm pro-
ductivity. Specifically, it is here analyzed the impact associated with the transfer of
knowledge through inter- firm labor mobility. In such perspective, technicians and high
skilled workers are viewed as knowledge carriers. Their role allows enterprises to ben-
efit from knowledge generated externally. However, knowledge transfers may not be
pure externalities if donor firms need to pay a wage premium to induce each inter-firm
flow. To assess these learning effects we implement both a regression and matching
approach. Our findings support the assumption that LbH enhances productivity at
firm level, confirming the importance of knowledge as production factor. In the re-
gression approach we estimate a parametric and semi-parametric production function
and evaluate the average contribution of LbH measures to firm TFP: elasticities vary
among industries and fluctuate mostly between 0.01 and 0.03. It turns to be extremely
interesting the evaluation of these learning effects among firms already performing such
recruitment strategy. Thus, by implementing a continuous treatment matching model
we find that treatment effects increase with doses of treatment. We find evidence that
firms focusing on such aspect of the recruitment may learn from external knowledge,
exploiting procedures and routines in the production processes previously matured in
24
other enterprises but similar environments. In fact, it seems that closeness matters
more than distance at least in the short run. Particular attention may be devoted not
only to the hiring of workers holding degrees in natural sciences, but also to those ones
characterized by degrees in other fields like social sciences and humanities or vocational
education. Likely the latter represent potential bridges for the exchange of quite tacit
or poorly articulated knowledge between firms. The implemented empirical analyses
shed lights on the existence of knowledge exchanges between any pair of enterprises,
hence it is not a phenomenon involving more knowledge based or larger firms only. As
a further argument, we briefly try to provide an economic qualification of our findings.
Looking at Table 4, the sum of the elasticities associated with LbH measures varies
between 0.027 and 0.025 in the estimation for the full sample. Assuming a 1 per cent
average increase in the LbH measures and given that in the time horizon taken into ac-
count (1995-2005) the average TFP growth has reached 0.3 percentage points (OECD
statistic tables), then it is possible to state that the overall contribution of LbH to TFP
growth would had been around 10 per cent. Although the assumption is not confirmed
by data, we still think that it may be a plausible firm strategy aimed at the increase of
its productivity. Moreover, it is worth underlining that Denmark experienced a rapid
increase in the TFP during the 90s with respect to several other developed countries,
thus the percentage contribution hypothetically imputed to LbH is not upward biased
by a particular level of TFP growth. That may partially confirm the importance of
any form of knowledge as production factor and provide some micro-founded empirical
arguments for NGT models.
25
References
[1] Adams, J.D. (1990). Fundamental Stocks of Knowledge and Productivity Growth.
Journal of Political Economy, 98(41), 673-702.
[2] Aghion, P. and Howitt, P. (1992). A Model of Growth through Creative Destruc-
tion. Econometrica, 60, 323-51.
[3] Almeida, P. and Kogut B. (1999). Localization of Knowledge and the Mobility of
Engineers in Regional Networks. Management Science 45, 905-916.
[4] Antonelli, C. (1995). The Economics of Localized Technological Change. Boston,
MA.
[5] Bia, M. and Mattei, A. (2008). A Stata package for the estimation of the dose-
response function through adjustment for the generalized propensity score. The
Stata Journal 8: 354-373.
[6] Cincera, M. (2005). Firms’ Productivity Growth and R&D Spillovers: An Analysis
of Alternative Technological Proximity Measures. CEPR Discussion Paper No.
4894.
[7] Gilson, R. J. (1999). The Legal Infrastructure of High Technology Industrial Dis-
tricts: Silicon Valley, Route 128, and Covenants not to Compete. New York Uni-
versity Law Review 74, 575-629.
[8] Grossman, G. M. and Helpman, E. (1991). Innovation and growth in the global
economy, Cambridge, MIT.
[9] Hirano, K. and Imbens, G. W. (2004). The Propensity Score with Continuous
Treatments, Working Paper, Department of Economics, University of California
at Berkeley.
26
[10] Inkmann, J. and Pohlmeier, W. (1995). R&D Spillovers, Technological Distance
and Innovative Success. Paper presented at the Conference on ’R&D, Innovation
and Productivity’, Institute for Fiscal Studies, London, May 15th-16th, 1995.
[11] Jaffe, A.B. (1986). Technological Opportunity and Spillovers of R&D. American
Economic Review, 76, 984-1001.
[12] Kaiser, U., Kongsted H. and Rnde T. (2008). Labor Mobility and Patenting Ac-
tivity. CAM working paper no. 2008-07.
[13] Kaiser, U., Licht G., Rnde T. and Schneider C. (2005). Patenting Activity in
Denmark. 2005-09, Copenhagen.
[14] Kim, J. and Marschke G. (2005). Labor mobility of scientists, technological dif-
fusion, and the firm’s patenting decision, RAND Journal of Economics, 36(2),
298-317.
[15] Levinsohn J. and Petrin A. (2000). Estimating Production Functions Using Inputs
to Control for Unobservables. Review of Economic Studies, Blackwell Publishing,
vol. 70(2), pages 317-341, 04.
[16] Maliranta, M., Mohnen, P. and Rouvinen P. (2009). Is Inter-Firm Labor Mobility
a Channel of Knowledge spillovers? Evidence from a Linked Employer-Employee
Panel. Industrial and Corporate Change, forthcoming.
[17] Olley, G. S. and Pakes, A. (1996). The dynamics of productivity in the telecomu-
nications equipment industry. Review of Economic Studies 69: 245-276.
[18] OECD (2005). Economic Survey of Denmark.
[19] Petit, M.L. and Tolwinski, B. (1996). Technology Sharing Cartels and Industrial
Structure. International Journal of Industrial Organization, 15, 77-101.
27
[20] Romer, P.M. (1990). Endogenous Technological Change. Journal of Political Econ-
omy, 98(5).
[21] Rosenkopf, L. and Almeida, P. (2003). Overcoming local search through alliances
and mobility. Management Science, 49(6): 751-766.
[22] Song, J., Almeida P., and Wu G. (2003). Learning-by-Hiring: When Is Mobility
More Likely to Facilitate Inter-firm Knowledge Transfer?. Management Science
49, 351-365.
[23] Yasar M., Raciborski R. and Poi B. (2008). Production function estimation in
Stata using the Olley and Pakes method. Stata Journal, StataCorp LP, vol. 8(2),
pages 221-231, June.
28
Tab
le1:
Des
crip
tive
stat
isti
cs
Tota
lSiz
e1
Siz
e2
Siz
e3
Acc
ounti
ng
Vari
able
s:M
edia
nM
ean
Sd
Media
nM
ean
Sd
Media
nM
ean
Sd
Media
nM
ean
Sd
Val
ue
add
ed15
0667
18.0
078
422.
2714
4131
61.6
511
197.
0931
055.
8339
303.
0286
691.
0294
697.
6821
4030
.50
6131
93.9
0M
ater
ials
2164
.76
1702
9.10
2622
78.8
020
59.0
280
93.4
660
985.
9152
998.
4410
7439
.341
9312
.817
2895
.852
7813
.30
2040
933
Cap
ital
2272
.00
2165
6.78
5855
64.1
021
5696
35.9
728
2841
.853
263.
512
6536
.194
4463
.519
1437
7286
20.6
4165
061.
0F
orei
gnow
ner
ship
00.
010.
490
0.00
0.07
00.
010.
070
0.01
0.07
Work
forc
eV
ari
able
s:N
um
ber
ofem
plo
yees
311
.63
108.
403
5.59
7.44
6668
.89
14.0
418
136
1.92
839.
73L
bHm
easu
res
ofth
em
ain
anal
ysis
:L
bH
,co
unt
mea
sure
00.
3110
5.06
00.
010.
080
0.13
0.73
021
.23
881.
40L
bH
,co
nca
vem
easu
re0
0.05
0.76
00.
010.
230
0.27
1.43
01.
545.
66L
bH
,p
roxim
ity
mea
sure
00.
030.
650
0.01
0.21
00.
251.
340
1.38
4.81
Fir
msp
ecifi
cch
arac
teri
stic
sof
empl
oyee
sm
ales
(men
asa
pro
por
tion
ofal
lem
plo
yees
)0
0.38
0.43
00.
340.
420
0.26
0.36
00.
250.
35ag
e1(e
mp
loye
esag
ed15
-28
asa
pro
por
tion
ofal
lem
plo
yees
)0.
070.
300.
370.
20.
350.
380.
480.
420.
290.
470.
410.
27ag
e2(e
mp
loye
esag
ed29
-36
asa
pro
por
tion
ofal
lem
plo
yees
)0.
060.
210.
290.
170.
240.
290.
210.
210.
080.
220.
220.
07ag
e3(e
mp
loye
esag
ed37
-47
asa
pro
por
tion
ofal
lem
plo
yees
)0.
100.
230.
300.
160.
240.
290.
220.
210.
100.
230.
220.
09ag
e4(e
mp
loye
esag
ed48
-65
asa
pro
por
tion
ofal
lem
plo
yees
)0
0.27
0.35
00.
170.
230.
040.
160.
200.
030.
160.
19sk
ill1
(em
plo
yees
wit
ha
tert
iary
edu
cati
onas
ap
rop
orti
onof
all
emp
loye
es)
00.
050.
170
0.04
0.15
00.
070.
130.
030.
160.
19sk
ill2
(em
plo
yees
wit
ha
seco
nd
ary/v
oca
tion
aled
uca
tion
asa
pro
por
tion
ofal
lem
plo
yees
)0
0.59
0.37
0.66
0.62
0.34
0.59
0.59
0.18
0.23
0.22
0.09
man
ager
(man
ager
sas
ap
rop
orti
onof
all
emp
loye
es)
00.
410.
450
0.02
0.08
00.
020.
030
0.01
0.03
mid
dle
man
ager
(mid
dle
man
ager
sas
ap
rop
orti
onof
all
emp
loye
es)
00.
020.
080
0.41
0.45
00.
271.
430
0.26
0.36
blu
eco
llar
(blu
eco
llar
sas
ap
rop
orti
onof
all
emp
loye
es)
00.
430.
501
0.51
0.49
00.
251.
341
0.63
0.48
Sect
ors
:M
anu
fact
uri
ng
00.
190.
380
0.18
0.38
00.
430.
490
0.52
0.49
Con
stru
ctio
n0
0.20
0.40
00.
200.
400
0.10
0.30
00.
070.
24W
hol
esal
etr
ade
00.
380.
490
0.39
0.49
00.
250.
430
0.20
0.40
Tra
nsp
ort
00.
060.
240
0.07
0.25
00.
060.
240
0.06
0.23
Fin
anci
al0
0.17
0.36
00.
160.
370
0.15
0.35
00.
160.
36O
bse
rvati
ons
6820
6466
1181
1132
095
63
Not
es:
Size
1:E
mpl
oyee
s≤
49;
Size
2:49≤
Em
ploy
ees≤
99;
Size
3:10
0≤
Em
ploy
ees≤
149;
Size
4:E
mpl
oyee
s≥
150.
29
Table 2: Main elasticities from Fixed Effects and Levinhson and Petrin estimations ofthe production function.
Food, beverages and tobacco Sale and repair of motor vehiclesFE LP FE LP
Log(Kit) 0.45*** 0.29*** Log(Kit) 0.47*** 0.41***Capital (0.00) (0.00) Capital (0.00) (0.00)Log(Lit) 0.18*** 0.26*** Log(Lit) 0.17*** 0.33***Labor (0.00) (0.00) Labor (0.00) (0.00)Observations 14866 14957 Observations 31150 31256
Textiles Wholesale tradeFE LP FE LP
Log(Kit) 0.49*** 0.44*** Log(Kit) 0.57*** 0.41***Capital (0.00) (0.00) Capital (0.00) (0.00)Log(Lit) 0.30*** 0.48*** Log(Lit) 0.21*** 0.33***Labor (0.00) (0.00) Labor (0.00) (0.00)Observations 7411 7413 Observations 75588 31256
Wood products Retail tradeFE LP FE LP
Log(Kit) 0.48*** 0.45*** Log(Kit) 0.47*** 0.55***Capital (0.00) (0.00) Capital (0.00) (0.00)Log(Lit) 0.26*** 0.33*** Log(Lit) 0.16*** 0.39***Labor (0.00) (0.00) Labor (0.00) (0.00)Observations 23033 23057 Observations 114233 75703
Chemicals Hotels and restaurantsFE LP FE LP
Log(Kit) 0.52*** 0.35*** Log(Kit) 0.35*** 0.44***Capital (0.00) (0.00) Capital (0.00) (0.00)Log(Lit) 0.26*** 0.50*** Log(Lit) 0.15*** 0.21***Labor (0.00) (0.00) Labor (0.00) (0.00)Observations 8296 8296 Observations 38052 114598Other non-metallic mineral products Transport
FE LP FE LPLog(Kit) 0.48*** 0.51*** Log(Kit) 0.43*** 0.32***Capital (0.00) (0.00) Capital (0.00) (0.00)Log(Lit) 0.26*** 0.28*** Log(Lit) 0.22*** 0.12***Labor (0.00) (0.00) Labor (0.00) (0.00)Observations 4333 4359 Observations 42375 38362
Basic metals Post and telecommunicationsFE LP FE LP
Log(Kit) 0.50*** 0.48*** Log(Kit) 0.35*** 0.36***Capital (0.00) (0.00) Capital (0.00) (0.00)Log(Lit) 0.26*** 0.34*** Log(Lit) 0.22*** 0.41***Labor (0.00) (0.00) Labor (0.00) (0.00)Observations 56603 56695 Observations 1388 43304
Furniture Financial intermediationFE LP FE LP
Log(Kit) 0.52*** 0.30*** Log(Kit) 0.48*** 0.48***Capital (0.00) (0.00) Capital (0.00) (0.00)Log(Lit) 0.26*** 0.49*** Log(Lit) 0.16*** 0.34***Labor (0.00) (0.00) Labor (0.00) (0.00)Observations 10709 10716 Observations 23785 23776
Construction Business activitiesFE LP FE LP
Log(Kit) 0.43*** 0.36*** Log(Kit) 0.43*** 0.35***Capital (0.00) (0.00) Capital (0.00) (0.00)Log(Lit) 0.28*** 0.37*** Log(Lit) 0.23*** 0.49***Labor (0.00) (0.00) Labor (0.00) (0.00)Observations 137971 138346 Observations 85023 85158
Notes: The dependent variable in all estimations is the log value added at firmlevel. All regressions include whether the firm is foreign-owned, shares of managers,middle managers, females, highly educated workers, technicians, age categories anda full set of year, size and county dummies. Estimated standard errors are shownin parentheses. Significance levels: ***1%, **5%, *10%. FE: Fixed Effects. LP:Levinhson and Petrin.
30
Table 3: Main elasticities from Olley and Peakes estimation of the production function.
Manufacturing Construction Ws and retail trade Transport Financial and business activiesLog(Kit) 0.37*** 0.39*** 0.42*** 0.18*** 0.17***Capital (0.00) (0.00) (0.00) (0.00) (0.00)Log(Lit) 0.45*** 0.49*** 0.40*** 0.54*** 0.54***Labor (0.00) (0.00) (0.00) (0.00) (0.00)Observations 50587 57490 94134 56302 99718
Notes: The dependent variable in all estimations is the log value added at firmlevel. All regressions include whether the firm is foreign-owned, shares of managers,middle managers, females, highly educated workers, technicians, age categories anda full set of year, size and county dummies. Estimated standard errors are shown inparentheses. Significance levels: ***1%, **5%, *10%.
Table 4: Estimated learning by hiring effects, main results.
Model1 Model2 Model3 Model4 Model5 Model6 Model7 Model8 Model9Log(Lbh) 0.019*** 0.009** 0.008***
(0.002) (0.003) (0.002)Log(Lbh(t-1)) 0.028*** 0.025*** 0.019***
(0.002) (0.003) (0.002)Log(LbH-concave) 0.016*** 0.008*** 0.010***
(0.001) (0.002) (0.001)Log(LbH-concave(t-1)) 0.020*** 0.019*** 0.015***
(0.002) (0.002) (0.002)Log(LbH-prox) 0.005** 0.004* 0.004*
(0.002) (0.002) (0.002)Log(LbH-prox(t-1)) 0.009** 0.010** 0.008**
(0.004) (0.005) (0.004)Observations 496167 496167 5930 254395 254395 3285 496167 496167 5930
Notes: The dependent variable is: i) in Model1, Model2 and Model3 is the TFPestimated in FE version; ii) in Model4, Model5 and Model6 is the TFP estimated inOP version; iii) in Model7, Model8 and Model9 is the TFP estimated in LP version.All regressions include whether the firm is foreign-owned, shares of managers, middlemanagers, females, highly educated workers, technicians, age categories and a full setof year, size, industry and county dummies. Estimated standard errors are shown inparentheses. Significance levels: ***1%, **5%, *10%.
31
Table 5: Estimated learning by hiring effects by industry.
Model1Manufacturing Construction Ws and retail trade Transport Financial and business activies
Log(Lbh) 0.009** 0.013** 0.023*** 0.017 0.021***(0.004) (0.005) (0.006) (0.012) (0.003)
Log(Lbh(t-1)) 0.012** 0.032*** 0.026*** 0.064*** 0.036***(0.004) (0.007) (0.006) (0.012) (0.003)
Observations 100348 104595 185895 31317 74012Manufacturing Construction Ws and retail trade Transport Financial and business activies
Log(Lbh-concave) 0.006** 0.018*** 0.003 0.027** 0.028***(0.002) (0.002) (0.004) (0.011) (0.003)
Log(Lbh-concave(t-1)) 0.004* 0.025*** 0.008** 0.034** 0.038***(0.002) (0.003) (0.004) (0.011) (0.003)
Observations 100348 104595 185895 31317 74012Model2
Manufacturing Construction Ws and retail trade Transport Financial and business activiesLog(Lbh) 0.009** 0.008 0.006 0.003 0.005
(0.003) (0.008) (0.005) (0.014) (0.005)Log(Lbh(t-1)) 0.015*** 0.031*** 0.010 0.033** 0.033***
(0.004) (0.006) (0.006) (0.016) (0.004)Observations 50275 56582 92368 16210 38960
Manufacturing Construction Ws and retail trade Transport Financial and business activiesLog(Lbh-concave) 0.006** 0.013*** 0.005 0.013 0.007*
(0.002) (0.003) (0.004) (0.013) (0.004)Log(Lbh-concave(t-1)) 0.009*** 0.023*** 0.014** 0.027** 0.030***
(0.002) (0.004) (0.004) (0.014) (0.004)Observations 50275 56582 92368 16210 38960
Model3Manufacturing Construction Ws and retail trade Transport Financial and business activies
Log(Lbh) 0.020** 0.018*** 0.021** 0.011 0.011***(0.007) (0.005) (0.008) (0.010) (0.003)
Log(Lbh(t-1)) 0.024** 0.039*** 0.023** 0.063*** 0.029***(0.007) (0.007) (0.007) (0.011) (0.003)
Observations 100348 104595 185895 31317 74012Manufacturing Construction Ws and retail trade Transport Financial and business activies
Log(Lbh-concave) 0.006** 0.023*** –0.005 0.021** 0.018***(0.002) (0.003) (0.005) (0.010) (0.003)
Log(Lbh-concave(t-1)) 0.002 0.030*** –0.000 0.033** 0.032***(0.004) (0.004) (0.005) (0.011) (0.003)
Observations 100348 104595 185895 31317 74012
Notes: The dependent variable in model1, model2 and model3 is the TFP estimatedrespectively from FE, OP and LP version. All regressions include whether the firmis foreign-owned, shares of managers, middle managers, females, highly educatedworkers, technicians, age categories and a full set of year, size and county dummies.Estimated standard errors are shown in parentheses. Significance levels: ***1%,**5%, *10%.
Table 6: Summary statistics of the LbH proximity measures, among ”treated” firms(pooled sample).
Pooled sampleMean Median Sd
LbH 8.81 2 25.72Observations 4031LbH-concave 3.48 1 7.87Observations 4031LbH-proximity 4.17 0.99 5.5Observations 600
32
Table 7: Estimated dose response and treatment functions, pooled sample. Outcome:log of TFP.
LbH, count measureTreatment Levels Mean Treatment effects, delta=10% Std Error Treatment effects, delta=20% Std Error Treatment effects, delta=40% Std Error0.1 3.6720.2 3.673 0.001 0.0070.3 3.680 0.007 0.007 0.008 0.0140.4 3.684 0.004 0.007 0.010 0.0140.5 3.695 0.011 0.008 0.015 0.015 0.023 0.0270.6 3.702 0.008 0.008 0.019 0.015 0.029 0.0290.7 3.728 0.025 0.009 0.033 0.016 0.048 0.0290.8 3.757 0.029 0.009 0.054 0.017 0.073 0.0320.9 3.787 0.030 0.010 0.059 0.018 0.092 0.0331 3.815 0.029 0.010 0.059 0.019 0.113 0.0351.1 3.841 0.026 0.012 0.054 0.019 0.113 0.0381.2 3.863 0.022 0.015 0.048 0.022 0.106 0.038
LbH, concave measureTreatment Levels Mean Treatment effects, delta=10% Std Error Treatment effects, delta=20% Std Error Treatment effects, delta=40% Std Error0.1 3.6510.2 3.655 0.004 0.0050.3 3.660 0.005 0.005 0.009 0.0100.4 3.666 0.006 0.005 0.011 0.0110.5 3.674 0.007 0.006 0.014 0.011 0.023 0.0180.6 3.682 0.009 0.006 0.016 0.012 0.028 0.0190.7 3.692 0.010 0.006 0.019 0.013 0.032 0.0200.8 3.704 0.011 0.007 0.021 0.013 0.037 0.0210.9 3.716 0.013 0.007 0.024 0.014 0.043 0.0221 3.731 0.014 0.008 0.027 0.014 0.048 0.0231.1 3.746 0.016 0.008 0.030 0.015 0.054 0.0251.2 3.764 0.018 0.009 0.033 0.016 0.060 0.026
LbH, proximity measureTreatment Levels Mean Treatment effects, delta=10% Std Error Treatment effects, delta=20% Std Error Treatment effects, delta=40% Std Error0.1 3.5760.2 3.589 0.013 0.0080.3 3.604 0.015 0.009 0.028 0.0170.4 3.621 0.017 0.010 0.032 0.0170.5 3.640 0.019 0.010 0.036 0.017 0.064 0.0330.6 3.661 0.021 0.011 0.040 0.018 0.072 0.0340.7 3.683 0.023 0.012 0.043 0.019 0.079 0.0350.8 3.708 0.025 0.013 0.047 0.020 0.087 0.0370.9 3.735 0.027 0.014 0.051 0.022 0.095 0.0391 3.763 0.029 0.014 0.055 0.023 0.103 0.0421.1 3.794 0.031 0.015 0.059 0.025 0.110 0.0451.2 3.826 0.033 0.016 0.063 0.026 0.118 0.048
Notes: TFP estimated from in LP version. Estimated bootstrapped standard errors,1000 replications. The matching variables include: capital stock, whether the firmis foreign-owned, shares of managers, middle managers, females, highly educatedworkers, technicians, age categories and a full set of year, size, industry and countydummies.. All these characteristics have been measure the year before treatmentoccurs.
Table 8: Descriptive statistics of the LbH measures used in the sensitivity analysis.
Mean Median SdLbH count measure without the wage condition 2.08 0 457.78LbH count measure with a wage increase condition of 10 per cent 0.21 0 70.39LbH count measure excluding the knowledge carriers of coming from a downscaling firm 0.26 0 83.92LbH count measure excluding the knowledge carriers with a tertairy education in Humanities and Social Sciences and technicians 0.07 0 20.89LbH count measure excluding the knowledge carriers with a period of unemployment 0.29 0 101.11LbH count measure excluding the knowledge carriers with a part-time job before moving 0.30 0 104.86
33
Tab
le9:
Est
imat
edle
arnin
gby
hir
ing
effec
ts,
sensi
tivit
yan
alysi
s.
Mod
el1
Mod
el2
Mod
el3
Mod
el4
Mod
el5
Mod
el6
Mod
el7
Mod
el8
Mod
el9
Mod
el1
0M
od
el1
1M
od
el1
2M
od
el1
3M
od
el1
4M
od
el1
5M
od
el1
6M
od
el1
7M
od
el1
8M
od
el1
9M
od
el2
0M
od
el2
1M
od
el2
2M
od
el2
3M
od
el2
4M
od
el2
5L
og(L
bh)
0.00
6**
0.00
7***
0.00
9***
0.00
9***
0.00
9**
0.00
7***
0.00
7***
0.00
8***
(0.0
02)
(0.0
02)
(0.0
02)
(0.0
02)
(0.0
04)
(0.0
01)
(0.0
02)
(0.0
02)
Log
(Lbh(t
-1))
0.01
6***
0.01
9***
0.02
0***
0.01
9***
0.01
7***
0.02
0***
0.01
9***
0.01
9***
(0.0
02)
(0.0
02)
(0.0
02)
(0.0
02)
(0.0
04)
(0.0
01)
(0.0
02)
(0.0
02)
Log
(LbH
-con
cave
)0.
008*
**0.
009*
**0.
010*
**0.
010*
**0.
011*
*0.
009*
**0.
006*
*0.
006*
*(0
.002
)(0
.001
)(0
.001
)(0
.001
)(0
.004
)(0
.001
)(0
.002
)(0
.002
)L
og(L
bH
-con
cave
(t-1
))0.
013*
**0.
015*
**0.
015*
**0.
015*
**0.
016*
**0.
018*
**0.
014*
**0.
014*
**(0
.002
)(0
.002
)(0
.002
)(0
.002
)(0
.005
)(0
.001
)(0
.002
)(0
.002
)L
og(L
bH
-pro
x)
0.00
4*0.
003*
0.00
5**
0.00
5**
0.00
6*0.
004*
0.00
9**
0.00
8**
(0.0
02)
(0.0
02)
(0.0
02)
(0.0
02)
(0.0
03)
(0.0
02)
(0.0
03)
(0.0
03)
Log
(LbH
-pro
x(t
-1))
0.00
7*0.
008*
*0.
009*
*0.
011*
*0.
011*
*0.
005*
0.00
9**
0.00
9**
(0.0
04)
(0.0
04)
(0.0
04)
(0.0
04)
(0.0
05)
(0.0
03)
(0.0
03)
(0.0
03)
Log
(LbH
-dis
t)0.
006
(0.0
06)
Log
(LbH
-dis
t(t-
1))
0.00
7(0
.005
)O
bse
rvat
ions
3193
2931
9329
4173
4947
9449
4794
5921
4961
6749
6167
5930
4961
6749
6167
5930
4961
6749
6167
5930
4961
6749
6167
5930
4961
6749
6167
5930
4961
6749
6167
5930
5931
Not
es:
The
depe
nden
tva
riab
leis
the
TF
Pob
tain
edfr
omL
Pes
tim
atio
n.M
odel
1-M
odel
3do
n’t
incl
ude
firm
sw
hose
acco
unti
ngva
riab
les
have
been
impu
ted;
Mod
el4-
Mod
el6
dono
tin
clud
efir
ms
wit
hin
flow
sof
high
skill
edw
orke
rsan
dte
chni
cian
sbe
low
(abo
ve)
the
5th
(95t
h)pe
rcen
tile
;M
odel
7-M
odel
9ex
clud
efr
omth
ele
arni
ngby
hiri
ngm
easu
res
the
know
ledg
eca
rrie
rsex
peri
enci
ngle
ssth
an10
per
cent
wag
ein
crea
seaf
ter
mov
ing;
Mod
el10
-Mod
el12
excl
ude
thos
eco
min
gfr
omfir
ms
dow
nsca
ling
the
labo
rfo
rce;
Mod
el13
-Mod
el15
excl
ude
thos
eha
ving
ate
rtia
ryed
ucat
ion
inH
uman
itie
s,So
cial
Scie
nces
and
Eco
nom
ics;
Mod
el16
-Mod
el18
excl
ude
the
know
ledg
eca
rrie
rsex
peri
enci
nga
wag
ein
crea
seaf
ter
mov
ing;
Mod
el19
-Mod
el21
excl
ude
the
know
ledg
eca
rrie
rsex
peri
enci
ngun
empl
oym
ent
befo
rem
ovin
g;M
odel
22-M
odel
24ex
clud
eth
ekn
owle
dge
carr
iers
expe
rien
cing
atr
ansi
tion
from
apa
rt-t
ime
job
toa
full-
tim
ejo
baf
ter
mov
ing;
Mod
el25
incl
ude
allt
heor
igin
alsa
mpl
ebu
tin
trod
uce
ane
wap
proa
chof
adju
stin
gth
ein
ter-
firm
labo
rm
obili
tyfo
rth
ete
chno
logi
cald
ista
nce.
All
regr
essi
ons
incl
ude
whe
ther
the
firm
isfo
reig
n-ow
ned,
shar
esof
man
ager
s,m
iddl
em
anag
ers,
fem
ales
,hi
ghly
educ
ated
wor
kers
,te
chni
cian
s,ag
eca
tego
ries
and
afu
llse
tof
year
,si
ze,
indu
stry
and
coun
tydu
mm
ies.
Sign
ifica
nce
leve
ls:
***1
%,
**5%
,*1
0%.
34
Department of Economics: Skriftserie/Working Paper: 2003: WP 03-1 Søren Harck: Er der nu en strukturelt bestemt langsigts-ledighed
i SMEC?: Phillipskurven i SMEC 99 vis-à-vis SMEC 94. ISSN 1397-4831.
WP 03-2 Beatrice Schindler Rangvid: Evaluating Private School Quality
in Denmark. ISSN 1397-4831. WP 03-3 Tor Eriksson: Managerial Pay and Executive Turnover in the
Czech and Slovak Republics. ISSN 1397-4831. WP 03-4 Michael Svarer and Mette Verner: Do Children Stabilize
Marriages? ISSN 1397-4831. WP 03-5 Christian Bjørnskov and Gert Tinggaard Svendsen: Measuring
social capital – Is there a single underlying explanation? ISSN 1397-4831.
WP 03-6 Vibeke Jakobsen and Nina Smith: The educational attainment of
the children of the Danish ‘guest worker’ immigrants. ISSN 1397-4831.
WP 03-7 Anders Poulsen: The Survival and Welfare Implications of
Altruism When Preferences are Endogenous. ISSN 1397-4831. WP 03-8 Helena Skyt Nielsen and Mette Verner: Why are Well-educated
Women not Full-timers? ISSN 1397-4831. WP 03-9 Anders Poulsen: On Efficiency, Tie-Breaking Rules and Role
Assignment Procedures in Evolutionary Bargaining. ISSN 1397-4831.
WP 03-10 Anders Poulsen and Gert Tinggaard Svendsen: Rise and Decline
of Social Capital – Excess Co-operation in the One-Shot Prisoner’s Dilemma
Game. ISSN 1397-4831.
WP 03-11 Nabanita Datta Gupta and Amaresh Dubey: Poverty and Fertility: An Instrumental Variables Analysis on Indian Micro Data. ISSN 1397-4831.
WP 03-12 Tor Eriksson: The Managerial Power Impact on Compensation –
Some Further Evidence. ISSN 1397-4831. WP 03-13 Christian Bjørnskov: Corruption and Social Capital. ISSN 1397-
4831. WP 03-14 Debashish Bhattacherjee: The Effects of Group Incentives in an
Indian Firm – Evidence from Payroll Data. ISSN 1397-4831.
WP 03-15 Tor Eriksson och Peter Jensen: Tidsbegränsade anställninger –
danska erfarenheter. ISSN 1397-4831. WP 03-16 Tom Coupé, Valérie Smeets and Frédéric Warzynski: Incentives,
Sorting and Productivity along the Career: Evidence from a Sample of Top Economists. ISSN 1397-4831.
WP 03-17 Jozef Koning, Patrick Van Cayseele and Frédéric Warzynski:
The Effects of Privatization and Competitive Pressure on Firms’ Price-Cost Margins: Micro Evidence from Emerging Economies. ISSN 1397-4831.
WP 03-18 Urs Steiner Brandt and Gert Tinggaard Svendsen: The coalition
of industrialists and environmentalists in the climate change issue. ISSN 1397-4831.
WP 03-19 Jan Bentzen: An empirical analysis of gasoline price
convergence for 20 OECD countries. ISSN 1397-4831. WP 03-20 Jan Bentzen and Valdemar Smith: Regional income convergence
in the Scandinavian countries. ISSN 1397-4831. WP 03-21 Gert Tinggaard Svendsen: Social Capital, Corruption and
Economic Growth: Eastern and Western Europe. ISSN 1397-4831.
WP 03-22 Jan Bentzen and Valdemar Smith: A Comparative Study of
Wine Auction Prices: Mouton Rothschild Premier Cru Classé. ISSN 1397-4831.
WP 03-23 Peter Guldager: Folkepensionisternes incitamenter til at arbejde. ISSN 1397-4831.
WP 03-24 Valérie Smeets and Frédéric Warzynski: Job Creation, Job
Destruction and Voting Behavior in Poland. ISSN 1397-4831. WP 03-25 Tom Coupé, Valérie Smeets and Frédéric Warzynski: Incentives
in Economic Departments: Testing Tournaments? ISSN 1397-4831.
WP 03-26 Erik Strøjer Madsen, Valdemar Smith and Mogens Dilling-
Hansen: Industrial clusters, firm location and productivity – Some empirical evidence for Danish firms. ISSN 1397-4831.
WP 03-27 Aycan Çelikaksoy, Helena Skyt Nielsen and Mette Verner:
Marriage Migration: Just another case of positive assortative matching? ISSN 1397-4831.
2004: WP 04-1 Elina Pylkkänen and Nina Smith: Career Interruptions due to
Parental Leave – A Comparative Study of Denmark and Sweden. ISSN 1397-4831.
WP 04-2 Urs Steiner Brandt and Gert Tinggaard Svendsen: Switch Point
and First-Mover Advantage: The Case of the Wind Turbine Industry. ISSN 1397-4831.
WP 04-3 Tor Eriksson and Jaime Ortega: The Adoption of Job Rotation:
Testing the Theories. ISSN 1397-4831. WP 04-4 Valérie Smeets: Are There Fast Tracks in Economic
Departments? Evidence from a Sample of Top Economists. ISSN 1397-4831.
WP 04-5 Karsten Bjerring Olsen, Rikke Ibsen and Niels Westergaard-
Nielsen: Does Outsourcing Create Unemployment? The Case of the Danish Textile and Clothing Industry. ISSN 1397-4831.
WP 04-6 Tor Eriksson and Johan Moritz Kuhn: Firm Spin-offs in
Denmark 1981-2000 – Patterns of Entry and Exit. ISSN 1397-4831.
WP 04-7 Mona Larsen and Nabanita Datta Gupta: The Impact of Health on Individual Retirement Plans: a Panel Analysis comparing Self-reported versus Diagnostic Measures. ISSN 1397-4831.
WP 04-8 Christian Bjørnskov: Inequality, Tolerance, and Growth. ISSN
1397-4831. WP 04-9 Christian Bjørnskov: Legal Quality, Inequality, and Tolerance.
ISSN 1397-4831. WP 04-10 Karsten Bjerring Olsen: Economic Cooperation and Social
Identity: Towards a Model of Economic Cross-Cultural Integration. ISSN 1397-4831.
WP 04-11 Iben Bolvig: Within- and between-firm mobility in the low-wage
labour market. ISSN 1397-4831. WP 04-12 Odile Poulsen and Gert Tinggaard Svendsen: Social Capital and
Market Centralisation: A Two-Sector Model. ISSN 1397-4831. WP 04-13 Aditya Goenka and Odile Poulsen: Factor Intensity Reversal and
Ergodic Chaos. ISSN 1397-4831. WP 04-14 Jan Bentzen and Valdemar Smith: Short-run and long-run
relationships in the consumption of alcohol in the Scandinavian countries.
ISBN 87-7882-010-3 (print); ISBN 87-7882-011-1 (online). WP 04-15 Jan Bentzen, Erik Strøjer Madsen, Valdemar Smith and Mogens
Dilling-Hansen: Persistence in Corporate Performance? Empirical Evidence from Panel Unit Root Tests.
ISBN 87-7882-012-X (print); ISBN 87-7882-013-8 (online). WP 04-16 Anders U. Poulsen and Jonathan H.W. Tan: Can Information
Backfire? Experimental Evidence from the Ultimatum Game. ISBN 87-7882-014-6 (print); ISBN 87-7882-015-4 (online). WP 04-17 Werner Roeger and Frédéric Warzynski: A Joint Estimation of
Price-Cost Margins and Sunk Capital: Theory and Evidence from the European Electricity Industry.
ISBN 87-7882-016-2 (print); ISBN 87-7882-017-0 (online).
WP 04-18 Nabanita Datta Gupta and Tor Eriksson: New workplace practices and the gender wage gap.
ISBN 87-7882-018-9 (print); ISBN 87-7882-019-7 (online). WP 04-19 Tor Eriksson and Axel Werwatz: The Prevalence of Internal
Labour Markets – New Evidence from Panel Data. ISBN 87-7882-020-0 (print); ISBN 87-7882-021-9 (online). WP 04-20 Anna Piil Damm and Michael Rosholm: Employment Effects of
Dispersal Policies on Refugee Immigrants: Empirical Evidence. ISBN 87-7882-022-7 (print); ISBN 87-7882-023-5 (online). 2005: WP 05-1 Anna Piil Damm and Michael Rosholm: Employment Effects of
Dispersal Policies on Refugee Immigrants: Theory. ISBN 87-7882-024-3 (print); ISBN 87-7882-025-1 (online). WP 05-2 Anna Piil Damm: Immigrants’ Location Preferences: Exploiting
a Natural Experiment. ISBN 87-7882-036-7 (print); ISBN 87-7882-037-5 (online). WP 05-3 Anna Piil Damm: The Danish Dispersal Policy on Refugee
Immigrants 1986-1998: A Natural Experiment? ISBN 87-7882-038-3 (print); ISBN 87-7882-039-1 (online). WP 05-4 Rikke Ibsen and Niels Westergaard-Nielsen: Job Creation and
Destruction over the Business Cycles and the Impact on Individual Job Flows in Denmark 1980-2001.
ISBN 87-7882-040-5 (print); ISBN 87-7882-041-3 (online). WP 05-5 Anna Maria Kossowska, Nina Smith, Valdemar Smith and Mette
Verner: Til gavn for bundlinjen – Forbedrer kvinder i topledelse og bestyrelse danske virksomheders bundlinje?
ISBN 87-7882-042-1 (print); ISBN 87-7882-043-X (online). WP 05-6 Odile Poulsen and Gert Tinggaard Svendsen: The Long and
Winding Road: Social Capital and Commuting. ISBN 87-7882-044-8 (print); ISBN 87-7882-045-6 (online). WP 05-7 Odile Poulsen and Gert Tinggaard Svendsen: Love Thy
Neighbor: Bonding versus Bridging Trust. ISBN 87-7882-062-6 (print); ISBN 87-7882-063-4 (online).
WP 05-8 Christian Bjørnskov: Political Ideology and Economic Freedom. ISBN 87-7882-064-2 (print); ISBN 87-7882-065-0 (online). WP 05-9 Sebastian Buhai and Coen Teulings: Tenure Profiles and
Efficient Separation in a Stochastic Productivity Model. ISBN 87-7882-066-9 (print); ISBN 87-7882-067-7 (online). WP 05-10 Christian Grund and Niels Westergård-Nielsen: Age Structure of
the Workforce and Firm Performance. ISBN 87-7882-068-5 (print); ISBN 87-7882-069-3 (online). WP 05-11 Søren Harck: AD-AS på dansk. ISBN 87-7882-070-7 (print); ISBN 87-7882-071-5 (online). WP 05-12 Søren Harck: Hviler Dansk Økonomi på en Cobb-Douglas
teknologi? ISBN 87-7882-092-8 (print); ISBN 87-7882-093-6 (online). 2006: WP 06-1 Nicolai Kristensen and Edvard Johansson: New Evidence on
Cross-Country Differences in Job Satisfaction Using Anchoring Vignettes.
ISBN 87-7882-094-4 (print); ISBN 87-7882-095-2 (online). WP 06-2 Christian Bjørnskov: How Does Social Trust Affect Economic
Growth? ISBN 87-7882-096-0 (print); ISBN 87-7882-097-9 (online). WP 06-3 Jan Bentzen, Erik Strøjer Madsen and Valdemar Smith: The
Growth Opportunities for SMEs? ISBN 87-7882-098-7 (print); ISBN 87-7882-099-5 (online). WP 06-4 Anna Piil Damm: Ethnic Enclaves and Immigrant Labour
Market Outcomes: Quasi-Experimental Evidence. ISBN 87-7882-100-2 (print); ISBN 87-7882-101-0 (online). WP 06-5 Svend Jespersen, Nicolai Kristensen og Lars Skipper: En kritik
af VEU-udvalgets arbejde. ISBN 87-7882-159-2 (print); ISBN 87-7882-160-6 (online). WP 06-6 Kræn Blume and Mette Verner: Welfare Dependency among
Danish Immigrants. ISBN 87-7882-161-4 (print); ISBN 87-7882-162-2 (online).
WP 06-7 Jürgen Bitzer, Wolfram Schrettl and Philipp J.H. Schröder:
Intrinsic Motivation versus Signaling in Open Source Software Development.
ISBN 87-7882-163-0 (print); ISBN 87-7882-164-9 (online). WP 06-8 Valérie Smeets, Kathryn Ierulli and Michael Gibbs: Mergers of
Equals & Unequals. ISBN 87-7882-165-7 (print); ISBN 87-7882-166-5 (online). WP 06-9 Valérie Smeets: Job Mobility and Wage Dynamics. ISBN 87-7882-167-3 (print); ISBN 87-7882-168-1 (online). WP 06-10 Valérie Smeets and Frédéric Warzynski: Testing Models of
Hierarchy: Span of Control, Compensation and Career Dynamics.
ISBN 87-7882-187-8 (print); ISBN 87-7882-188-6 (online). WP 06-11 Sebastian Buhai and Marco van der Leij: A Social Network
Analysis of Occupational Segregation. ISBN 87-7882-189-4 (print); ISBN 87-7882-190-8 (online). 2007: WP 07-1 Christina Bjerg, Christian Bjørnskov and Anne Holm: Growth,
Debt Burdens and Alleviating Effects of Foreign Aid in Least Developed Countries.
ISBN 87-7882-191-6 (print); ISBN 87-7882-192-4 (online). WP 07-2 Jeremy T. Fox and Valérie Smeets: Do Input Quality and
Structural Productivity Estimates Drive Measured Differences in Firm Productivity?
ISBN 87-7882-193-2 (print); ISBN 87-7882-194-0 (online). WP 07-3 Elisabetta Trevisan: Job Security and New Restrictive
Permanent Contracts. Are Spanish Workers More Worried of Losing Their Job?
ISBN 87-7882-195-9 (print); ISBN 87-7882-196-7 (online). WP 07-4 Tor Eriksson and Jaime Ortega: Performance Pay and the “Time
Squeeze”. ISBN 9788778822079 (print); ISBN 9788778822086 (online).
WP 07-5 Johan Moritz Kuhn: My Pay is Too Bad (I Quit). Your Pay is Too Good (You’re Fired).
ISBN 9788778822093 (print); ISBN 9788778822109 (online). WP 07-6 Christian Bjørnskov: Social trust and the growth of schooling. ISBN 9788778822116 (print); ISBN 9788778822123 (online). WP 07-7 Jan Bentzen and Valdemar Smith: Explaining champagne prices
in Scandinavia – what is the best predictor? ISBN 9788778822130 (print); ISBN 9788778822147 (online). WP 07-8 Sandra Cavaco, Jean-Michel Etienne and Ali Skalli: Identifying
causal paths between health and socio-economic status: Evidence from European older workforce surveys
ISBN 9788778822154 (print); ISBN 9788778822161 (online). WP 07-9 Søren Harck: Long-run properties of some Danish macro-
econometric models: an analytical approach. ISBN 9788778822390 (print); ISBN 9788778822406 (online). WP 07-10 Takao Kato and Hideo Owan: Market Characteristics, Intra-Firm
Coordination, and the Choice of Human Resource Management Systems: Evidence from New Japanese Data.
ISBN 9788778822413 (print); ISBN 9788778822420 (online). WP 07-11 Astrid Würtz: The Long-Term Effect on Children of Increasing
the Length of Parents’ Birth-Related Leave. ISBN 9788778822437 (print); ISBN 9788778822444 (online). WP 07-12 Tor Eriksson and Marie-Claire Villeval: Performance Pay,
Sorting and Social Motivation. ISBN 9788778822451 (print); ISBN 9788778822468 (online). WP 07-13 Jane Greve: Obesity and Labor Market Outcomes: New Danish
Evidence. ISBN 9788778822475 (print); ISBN 9788778822482 (online). 2008: WP 08-1 Sebastian Buhai, Miguel Portela, Coen Teulings and Aico van
Vuuren: Returns to Tenure or Seniority ISBN 9788778822826 (print); ISBN 9788778822833 (online).
WP 08-2 Flora Bellone, Patrick Musso, Lionel Nesta et Frédéric Warzynski: L’effet pro-concurrentiel de l’intégration européenne : une analyse de l’évolution des taux de marge dans les industries manufacturières françaises
ISBN 9788778822857 (print); ISBN 9788778822864 (online). WP 08-3 Erdal Yalcin: The Proximity-Concentration Trade-Off under
Goods Price and Exchange Rate Uncertainty ISBN 9788778822871 (print); ISBN 9788778822888 (online) WP 08-4 Elke J. Jahn and Herbert Brücker: Migration and the Wage
Curve: A Structural Approach to Measure the Wage and Employment Effects of Migration
ISBN 9788778822895 (print); ISBN 9788778822901 (online) WP 08-5 Søren Harck: A Phillips curve interpretation of error-correction
models of the wage and price dynamics ISBN 9788778822918 (print); ISBN 9788778822925 (online) WP 08-6 Elke J. Jahn and Thomas Wagner: Job Security as an
Endogenous Job Characteristic ISBN 9788778823182 (print); ISBN 9788778823199 (online) WP 08-7 Jørgen Drud Hansen, Virmantas Kvedaras and Jørgen Ulff-
Møller Nielsen: Monopolistic Competition, International Trade and Firm Heterogeneity - a Life Cycle Perspective -
ISBN 9788778823212 (print); ISBN 9788778823229 (online) WP 08-8 Dario Pozzoli: The Transition to Work for Italian University
Graduates ISBN 9788778823236 (print); ISBN 9788778823243 (online)
WP 08-9 Annalisa Cristini and Dario Pozzoli: New Workplace Practices
and Firm Performance: a Comparative Study of Italy and Britain ISBN 9788778823250 (print); ISBN 9788778823267 (online)
WP 08-10 Paolo Buonanno and Dario Pozzoli: Early Labour Market
Returns to College Subjects ISBN 9788778823274 (print); ISBN 9788778823281 (online)
WP 08-11 Iben Bolvig: Low wage after unemployment - the effect of
changes in the UI system ISBN 9788778823441 (print); ISBN 9788778823458 (online)
WP 08-12 Nina Smith, Valdemar Smith and Mette Verner: Women in Top Management and Firm Performance ISBN 9788778823465 (print); ISBN 9788778823472 (online)
WP 08-13 Sebastian Buhai, Elena Cottini and Niels Westergård-Nielsen:
The impact of workplace conditions on firm performance ISBN 9788778823496 (print); ISBN 9788778823502 (online)
WP 08-14 Michael Rosholm: Experimental Evidence on the Nature of the
Danish Employment Miracle ISBN 9788778823526 (print); ISBN 9788778823533 (online)
WP 08-15 Christian Bjørnskov and Peter Kurrild-Klitgaard: Economic
Growth and Institutional Reform in Modern Monarchies and Republics: A Historical Cross-Country Perspective 1820-2000
ISBN 9788778823540 (print); ISBN 9788778823557 (online)
WP 08-16 Nabanita Datta Gupta, Nicolai Kristensen and Dario Pozzoli: The Validity of Vignettes in Cross-Country Health Studies
ISBN 9788778823694 (print); ISBN 9788778823700 (online) WP 08-17 Anna Piil Damm and Marie Louise Schultz-Nielsen: The
Construction of Neighbourhoods and its Relevance for the Measurement of Social and Ethnic Segregation: Evidence from Denmark ISBN 9788778823717 (print); ISBN 9788778823724 (online)
WP 08-18 Jørgen Drud Hansen and Jørgen Ulff-Møller Nielsen: Price as an
Indicator for Quality in International Trade? ISBN 9788778823731 (print); ISBN 9788778823748 (online)
WP 08-19 Elke J. Jahn and Thomas Wagner: Do Targeted Hiring Subsidies
and Profiling Techniques Reduce Unemployment? ISBN 9788778823755 (print); ISBN 9788778823762 (online)
WP 08-20 Flora Bellone, Patrick Musso, Lionel Nesta and Frederic
Warzynski: Endogenous Markups, Firm Productivity and International Trade: Testing Some Micro-Level Implications of the Melitz-Ottaviano Model ISBN 9788778823779 (print); ISBN 9788778823786 (online)
WP 08-21 Linda Bell, Nina Smith, Valdemar Smith and Mette Verner: Gender differences in promotion into top-management jobs ISBN 9788778823830 (print); ISBN 9788778823847(online)
WP 08-22 Jan Bentzen and Valdemar Smith: An empirical analysis of the
relationship between the consumption of alcohol and liver cirrhosis mortality ISBN 9788778823854 (print); ISBN 9788778823861(online)
WP 08-23 Gabriel J. Felbermayr, Sanne Hiller and Davide Sala: Does
Immigration Boost Per Capita Income? ISBN 9788778823878 (print); ISBN 9788778823885(online)
WP 08-24 Christian Gormsen: Anti-Dumping with Heterogeneous Firms:
New Protectionism for the New-New Trade Theory ISBN 9788778823892 (print); ISBN 9788778823908 (online)
WP 08-25 Andrew E. Clark, Nicolai Kristensen
and Niels Westergård-
Nielsen: Economic Satisfaction and Income Rank in Small Neighbourhoods ISBN 9788778823915 (print); ISBN 9788778823922 (online)
WP 08-26 Erik Strøjer Madsen and Valdemar Smith: Commercialization of Innovations and Firm Performance ISBN 9788778823939 (print); ISBN 9788778823946 (online)
WP 08-27 Louise Lykke Brix and Jan Bentzen: Waste Generation In
Denmark 1994-2005 An Environmental And Economic Analysis
ISBN 9788778823953 (print); ISBN 9788778823977 (online) WP 08-28 Ingo Geishecker, Jørgen Ulff-Møller Nielsen and Konrad
Pawlik: How Important is Export-Platform FDI? Evidence from Multinational Activities in Poland ISBN 9788778823984 (print); ISBN 9788778823991 (online)
WP 08-29 Peder J. Pedersen and Mariola Pytlikova: EU Enlargement:
Migration flows from Central and Eastern Europe into the Nordic countries - exploiting a natural experiment ISBN 9788778824004 (print); ISBN 9788778824028 (online)
2009: WP 09-1 Tomi Kyyrä, Pierpaolo Parrotta and Michael Rosholm:
The Effect of Receiving Supplementary UI Benefits on Unemployment Duration ISBN 9788778824035 (print); ISBN 9788778824042 (online)
WP 09-2 Dario Pozzoli and Marco Ranzani: Old European Couples’
Retirement Decisions: the Role of Love and Money ISBN 9788778824165 (print); ISBN 9788778824172 (online)
WP 09-3 Michael Gibbs, Mikel Tapia and Frederic Warzynski: Globalization, Superstars, and the Importance of Reputation: Theory & Evidence from the Wine Industry
ISBN 9788778824189 (print); ISBN 9788778824196 (online) WP 09-4 Jan De Loecker and Frederic Warzynski: Markups and Firm-
Level Export Status ISBN 9788778824202 (print); ISBN 9788778824219 (online) WP 09-5 Tor Eriksson, Mariola Pytliková and Frédéric Warzynski:
Increased Sorting and Wage Inequality in the Czech Republic: New Evidence Using Linked Employer-Employee Dataset
ISBN 9788778824226 (print); ISBN 9788778824233 (online) WP 09-6 Longhwa Chen and Tor Eriksson: Vacancy Duration, Wage
Offers, and Job Requirements – Pre-Match Data Evidence ISBN 9788778824240 (print); ISBN 9788778824257 (online) WP 09-7 Tor Eriksson, Valérie Smeets and Frédéric Warzynski: Small
Open Economy Firms in International Trade: Evidence from Danish Transactions-Level Data
ISBN 9788778823861 (print); ISBN 9788778823878 (online) WP 09-8 Dario Pozzoli and Marco Ranzani: Participation and Sector
Selection in Nicaragua ISBN 9788778823885 (print); ISBN 9788778823892 (online) WP 09-9 Rikke Ibsen, Frederic Warzynski and Niels Westergård-Nielsen:
Employment Growth and International Trade: A Small Open Economy Perspective
ISBN 9788778823908 (print); ISBN 9788778823915 (online)
WP 09-10 Roger Bandick and Holger Görg: Foreign acquisition, plant survival, and employment growth ISBN 9788778823922 (print); ISBN 9788778823939 (online) WP 09-11 Pierpaolo Parrotta and Dario Pozzoli: The Effect of Learning by Hiring on Productivity ISBN 9788778823946 (print); ISBN 9788778823953 (online) WP 09-12 Takao Kato and Pian Shu Peer Effects, Social Networks, and Intergroup Competition in the Workplace ISBN 9788778823984 (print); ISBN 9788778823991 (online) WP 09-13 Sanne Hiller and Erdal Yalcin: Switching between Domestic Market Activity, Export and FDI ISBN 9788778824004 (print); ISBN 9788778824028 (online) WP 09-14 Tor Eriksson and Mariola Pytlikova: Foreign Ownership Wage Premia in Emerging Economies: Evidence from Czech Republic ISBN 9788778824035 (print); ISBN 9788778824042 (online) WP 09-15 Astrid Würtz Rasmussen: Family Structure Changes and Children´s Health, Behavior, and Educational Outcomes ISBN 9788778824059 (print); ISBN 9788778824066 (online) WP 09-16 Tor Eriksson: How Many Danish Jobs Can (Potentially) Be Done Elsewhere? ISBN 9788778824073 (print); ISBN 9788778824080 (online) WP 09-17 Lorenzo Cappellari, Claudio Lucifora and Dario Pozzoli: Determinants of Grades in Maths for Students in Economics ISBN 9788778824103 (print); ISBN 9788778824110 (online)