the effect of energy conservation tax credits on minority household housing improvements

12

Click here to load reader

Upload: martin-williams

Post on 19-Aug-2016

215 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: The effect of energy conservation tax credits on minority household housing improvements

T H E E F F E C T O F E N E R G Y C O N S E R V A T I O N T A X C R E D I T S ON M I N O R I T Y H O U S E H O L D H O U S I N G

I M P R O V E M E N T S

Martin Williams and David Poyer

INTRODUCTION

This article examines whether the state and federal tax credits allowed to households affect their energy-conservation improvements activity. Tax credits, in general, are a lively public policy instrument in the cur- rent political environment. Yet virtually nothing is known about the im- pact of the energy tax credit on the discrete decisions about insulation improvements made by black and Hispanic households. Walsh (1989), for example, has analyzed the effects of energy tax credits, fuel prices, housing-specific characteristics, weather and an assorted array of demo- graphic variables on energy conservation activity for all types of house- holds combined. 1 Tax credits have virtually no impact on the household's decision to make energy improvements. The earlier results are from a sample of households made up of renters residing in multifamily units and from single family owner-occupied units. But most renters hardly make insulation investment decisions on the units they rent. Furthermore some homeowners are restricted in their residential location search to find new housing stock, and it is quite possible that they may react differently to any incentives to make energy-conservation improvements of the oldest existing housing stock. Until now, empirical analysis of energy-conservation improvement activity has not been focusing on house- holds broken down by minority type. It is this group that is more likely to be restricted in their new residential location search.

This article disaggregates the household sample by ethnicity, taking into account only those black and Hispanic households who are homeowners. We examine the influence of tax credits, fuel prices, exter- nal temperature conditions, and an array of housing characteristics and demographic variables on conservation improvement activity by these

Page 2: The effect of energy conservation tax credits on minority household housing improvements

124 The Review of Black Political Economy/Spring 1996

homeowner types. The remainder of the article is organized as follows: first we review the earlier models on energy conservation improvement activity; then we discuss the data and present the empirical results; fi- nally we summarize our findings and offer more suggestions for future work.

MODELING HOUSEHOLD INSULATION CHOICE

The past literature devoted to analyzing energy-saving home improve- ment choice behavior can be classified into two types. The first type develops generalized telephone survey information to evaluate the merits of the Residential Energy Conservation Tax Credit program on consumer decisions to insulate their homes 2 (see, e.g., Pitts and Wittenbach, 1981).

The second type develops empirical logit and probit models from a more direct connection to a specific utility model of household behavior. The consumers are assumed to maximize their utility by considering the quantity of energy conservation improvements and future period tem- perature control. They determine the optimum amount of energy-saving capital improvements by comparing the current costs of conservation investment against the potential future benefits from the investment (Walsh, 1989). 3

Given that we do not know the reliability of telephone survey re- sponses as a measure of consumer perceptions to the energy tax credit as a price reduction mechanism, the empirical results from the first type of studies cannot be easily interpreted relative to the more current work. Moreover, their usefulness for direct policy intervention may also be limited. Obviously, the distinction between perceptions typical in tele- phone survey responses and actual market adjustments is important.

The more recent analyses attempt to avoid this limitation. Walsh (1989) uses household data from the National Residential Energy Consumption survey to evaluate the impact of the tax credit on energy-conservation improvement choices. 4 He develops logit and probit models where the determinants of choice include external temperature conditions, future energy prices, housing structure characteristics, housing tenure of the occupant, net-tax price measure of energy saving equipment, along with the socioeconomic characteristics of the household. While Walsh exam- ines a set of factors that is more systematically linked to household improvement activity, he assumes that both renters and owners are equally informed of the tax credit and both face the same pretax prices. Natu- rally, renters do not make decisions on energy conservation require-

Page 3: The effect of energy conservation tax credits on minority household housing improvements

Williams and Poyer 125

ments; their landlords do. One way to avoid this problem is to consider a sample of exclusively homeowners. Also, there is some evidence that tax credits benefit higher income families more than lower-income families (Pitts and Wittenbach, 1981). s Unfortunately, this work did not assess the impact of the tax credit for minority households in evaluating its effect by income level. Moreover, it should be noted that the results of the recent study by Walsh have indicated that tax credits do not stimulate adoption of residential energy conservation capital improvements. One reason for this result may be that the measure of the price of energy-saving capital that is used in the study is inappropriate. Walsh assumes a con- stant pretax price (unity) of energy capital across geographic regions. But the more recent empirical evidence shows that capital equipment prices vary dramatically across regions (Garofalo and Molhatra, 1991). 6

In order to overcome some of the limitations that we point out in the earlier work, we examine the factors that are systematically associated with home insulation activity for Hispanic households and black house- holds respectively. We begin with a simple statement of a household making intertemporal decisions on the quantity of energy-saving equip- ment and control of future indoor temperatures that maximizes utility. This framework for our empirical models is adopted from Walsh (1989). 7 Note that the connection between the theory and the empirical estimates will not be direct, so we report two different model specifications and two estimators.

Implicitly, we accept Walsh's two-period decision process for a utility maximizing household. Assume a Cobb-Douglas utility function:

U=(Zo-R*PR,)ao+D[Z1-PfC(T1, R1, E)]~,T~ 2 (1)

where Z0 R*PR,

D Zl Pf C T1 R1 E G2

= current period gross income = current period expenditure on thermal integrity improve-

ment = a discount factor = future period gross income = future period fuel price = future period energy consumption = future period temperature control = future period thermal integrity = external temperature conditions = 1 - a 1.

Note that the first term in equation (1) can be interpreted in terms of

Page 4: The effect of energy conservation tax credits on minority household housing improvements

126 The Review of Black Political Economy/Spring 1996

current period income allocated to other goods after expenditures on the initial thermal integrity improvements are accounted for. The second term can be interpreted as future period income allocated to other goods net expenditures on future fuel consumption. Future energy consumption depends on the future thermal integrity of the house (RI), temperature control (T1), and external temperature conditions (E) respectively. Now R 1 depends on the initial thermal quality of the building (R0) and any changes in the house's initial thermal integrity (R*). So that R1 = R* + R0. Rewriting (1) we get:

U = ( Z o - R * PR,)oo + D [ Z i - P f C ( T I , R * Ro, E)]o~T72 (2)

The household chooses R* to maximize U subject to a budget constraint that is implied in the objective function (2). We set the partial derivative of (2) with respect to R* equal to zero. This gives us:

~U

~R* ~ = o%(Z0 -R*PR, )ao- ' ( -Pm)

+oq [ZI - PfC(TI, R * Ro, E)] r ( - P f )T~ 2 = 0 (3)

From (3) we express the optimum thermal improvements (R*) as:

R* = R(Zo, Ro, P* E, PI, ao, al, a2, Z1, T1) (4)

The optimum amount of thermal improvements will increase as the net price for such improvements falls. But the net price falls with the energy tax credit. So R P* becomes PR** = P~ (1-tp) here p is the proportion of expenditures on thermal improvements allowed to be assessed a tax credit. Also, the future period temperature control is affected by the area of the house (A). And a0, am, and a2 are utility parameters.

We expect that households with higher incomes (Z0) will increase energy-conservation improvements. The effect of future fuel price (Pf), area of home (A) and external temperatures (E) are ambiguous (see Walsh, 1989). What effects Pf, A, and E have on energy-saving improvements is an empirical question.

Thus the two-period Cobb-Douglas utility framework permits one to formulate the household choice problem so that it gives testable hypoth- eses about the relationship between economic and policy variables (tax credits) and energy-saving capital improvements.

Page 5: The effect of energy conservation tax credits on minority household housing improvements

Williams and Poyer 127

THE DATA AND EMPIRICAL MODELS

Our empirical analysis is based on the 1987 Residential Energy Con- sumption Survey (RESC) collected by the U.S. Department of Energy. 8 The data provide socioeconomic, demographic and energy-saving infor- mation. The data afford the possibility of a relatively detailed study of factors determining energy-conservation choice broken down by race. We consider information on black and Hispanic households who owned their own homes. All renters are excluded from the sample. Specifically, we report findings for a sample of 297 blacks and 168 Hispanics respec- tively. 9 Each individual in the sample made a yes or no decision to make wall and attic insulations in their homes. Thus choice is dichotomous.

The sample also identifies the age of household head, family income together with the area of the house to be heated, and the year the struc- ture was built. Moreover, the availability of data on expenditures and consumption of fuel permitted us to supplement each observation with measured prices of fuel. Information on the number of cooling degree days and the number of heating degree days is also available.

In the absence of specific data on the price of energy saving capital, we use state-by-state data on the price of private capital for 1985 as a proxy measure. These estimates are obtained from Garofalo and Molhatra (1991). l~ The price of capital is similar to that described by Field and Grebenstein (1988).11 The price is the weighted average of the price of machinery and buildings with the proportion of capital in each as weights. The formula accounts for the federal and state corporate tax rate, depre- ciation and the effective property tax rate in the state. The price of capital varies by state due to variation in state taxes. Here we are assuming that the price of private capital is a proxy price for energy-saving capital. Note also that this is an improvement over the measure used by Walsh, Actually, he assumed that the price of energy conservation equipment was constant across regions.

We obtain the energy tax credit information from Walsh (1989). They range from 15 percent of expenditures for the states in the northeast to 36 percent for states in the west. The equations used for estimating house- hold energy-conservation choice can be written in general form as in (5):

Z i ---- f(Hj, Ok, N) (5)

where Zi = the probability of making an energy-conservation decision, i; Dk is the kth household characteristics, Hj is the jth housing structure characteristic, and N captures the effect of other factors. The form of the

Page 6: The effect of energy conservation tax credits on minority household housing improvements

128 The Review of Black Political Economy/Spring 1996

empirical model has little a priori basis in theory. Previous studies have included different measures of fuel prices and demographic factors in explaining energy-conservation activity. The specific criteria used to de- rive the form of the empirical models are not straightforward. Cameron (1985) uses a translog specification of an indirect utility function, t2 Here the link between the theory and the form of the empirical model is direct. The variables enter as quadratic terms of prices and income together with interaction terms between income and prices. However, the specification does not control for other important demographic factors such as age, etc. But in Walsh (1989), the link between the theory and the estimates is not direct.

It is interesting that the literature on pretesting estimation procedures does not offer any clear-cut guidelines for model selection problems in the conventional linear model 13 (see Wallace, 1977). Thus we offer a brief justification for the procedures used to derive the final equations. Our practice was to the prior information on the variables and specifica- tions adopted in earlier work and the reasons for them. We observe the different results using agreement of the signs of the estimated coeffi- cients with a priori hypotheses, overall goodness of fit of the model, statistical significance of the measured effects and consistency with ear- lier work. In so doing, equation (6) reports the specification for the energy-conservation choice equation.

Z = a 0 + a~ PE + a2 INC + a 3 AGE + a4 HDD + a5 CDD + a6 AREA + a 7 NTP + a 8 H1-DUM + a 9 H2-DUM + al0 AGH + E (6)

where Z = dummy variable, which equals one if the choice is to insulate the attic and wall of the home and zero otherwise; PE = average price of energy for 1981 and 1987; INC = household family income; AGE = age of head of household; HDD = number of heating degree days; CDD = number of cooling degree days; AREA = square foot heated area of house; NTP = net-tax price of energy conservation capital; H1-DUM = dummy variable, which equals one if house was built before 1940 and zero otherwise; H2-DUM = dummy variable, which equals one if house was built between 1940-1970 and zero otherwise; AGH = measure of the interaction of age and H1-DUM; e, ~t = stochastic errors.

E M P I R I C A L R E S U L T S

As noted at the outset there are alternative methods for estimation of dichotomous-dependent A variable models. The most conventional speci-

Page 7: The effect of energy conservation tax credits on minority household housing improvements

Williams and Poyer 129

fication is a linear probability functional form. However, it has been noted in the literature 14 (see, e.g., Golberger, 1964) that the linear model has together with the use of ordinary least squares (OLS) to estimate it, major problems, that are noted as nonlinear distribution of the error term; inherent heteroskedastic disturbances; predictions that are not limited to zero or unity, and an adjusted goodness of fit that is not an accurate measure of overall fit. Furthermore, we have no guidance from economic theory to select one estimation method (and therefore, implicitly a model) for the dichotomous dependent variable problem. In large part, previous work has relied upon the judgment of individual researchers in making a choice.

Since the comparative evaluation of the effects of the choice of an estimator with specific samples 15 (Gunderson, 1974) indicates little ap- preciable effects of individual techniques on the results, we estimate equation (6) using logit and probit methods. Equations (7) and (8) com- pare the probit and logit models specified for each where E( ) is the expected value:

E(ZI X) = (1 + exp(-XtB)) -1 (logit) (7)

~.XLSexP2 2dt (probit) E(ZIX)-.~ ~ -a (8)

where Z = binary dependent variable; X t = Ix k vector of independent variables; B = K x 1 parameter vector.

Table 1 presents the estimates for equation (6) for black households. The two specifications (models 1 and 2) are differentiated by tests of the measured effect of the interaction between age and housing built before 1940. Several patterns in the estimated coefficients are of interest.

We find that the socioeconomic related variables income (INC) and age (AGE) are consistently significant determinants of energy conserva- tion choice among black households. Older people are less likely to install conservation equipment. But higher income households are more likely to make the decision to install. One other way to think of the effect of income is that higher income black households are more likely to be aware of the effective reduction in the price of energy-saving capital. ~6 Future energy prices (PE) are not important consideration in the decision to insulate the home. The effect of external temperature conditions on the thermal integrity of the house are captured by HDD and CDD respec- tively. HDD measures the daily shortfall of outside temperatures below

Page 8: The effect of energy conservation tax credits on minority household housing improvements

130 The Review of Black Political Economy/Spring 1996

TABLE 1 Estimates of Energy-saving Housing Improvements:

(Black Households)

Independem VaHable~

InlefcepI

PE

INC

AGE

ilDD

CDD

AREA (r.q. fl)

NTP

III.DUM (< 1940)

112-DUM ( 1940-1970)

AGII

Medici I probit

Coefr. i.vlllue

2.2799 1.82

0,2500 0.56

0.00~I 1.63"

-0.0117 -I.93"

0.00007 0.79

0.0C027 1.78-

0,00023 1.52"

-I0.S070 -2.4O"

-0.97111 -3.15"

-0.4228 -I.47-

Loeit

Coeff. t-value

3.61114 1.70

053617 I 0.73

O.(Xl(~2 I .$6"

�9 0.0205 -2.00"

0 . ~ 1 0.74

0.0005 1.72-

0.0004 1.52"

-17.5560 -2.29"

-I.6889 -3.04"

-0.74911 - I .43"

Pmbil

Ceeff. I-value~

1,9390 1.54

0.2.329 0.52

0.0001 1.63-

-0.0117 -I.60"

O.(XXX~ 0.68

0.0027 1.76"

0.00022 1.48"

-10.2380 -2.35"

�9 0.3701 -0.$5

-0.0044 -0.311

297 No. OB$ 297 297

Adj. R 2 . . . . . . . . . . . .

X z 49.20 (900

49.00 (9dO

46.85 (gdl)

~,tr II l..neil

Coeff. t-vtfue~

3.0735 1.41

0.5360 0.70

o.(moo'2 1.37"

-0.0206 - 1.64-

0.00009 0.66

0.0C046 1.75-

0.0004 1.52"

-17.41~ -2.27"

�9 0.7617 -0.67

-0.0031 -0.26

297

46.0g (9d0

It should be noted that the t-statistics are based on asymptotic distribution theory for probit and logit. As such we do not make comparisons across estimators. The X2(df) is the chi-square statistic for the overall association with maximum likelihood methods. *These parameter estimates are significant at the 95 percent level of confidence. **These parameter estimates are significant at the 90 percent level of confidence.

an established reference level cumulated annually. The findings indicate that HDD does not exert a significant influence. On the other hand, CDD measures the daily shortfall of outside temperature above an established reference level cumulated annually. This last factor is a statistically sig- nificant determinant of the decision to insulate the home.

Next we examine the factors that capture the structural characteristics of the house. We find that the area of the house to be heated (AREA) has a statistically positive effect on choice. On the other hand the age of the structure (H1-DUM) and (H2-DUM) respectively have a statistically sig- nificant negative influence.

The net-tax price (NTP) measure of energy conservation capital ap- pears to exert a significant influence on the decision to insulate the home.

Page 9: The effect of energy conservation tax credits on minority household housing improvements

Williams and Poyer 131

TABLE 2 Estimates of Energy-saving Housing Improvements:

(Hispanic Households)

Independent Variables

Inlercepl

PE

INC

AGE

IIDD

CDD

AREA (~q. fl)

NTP .29.742

i l I -DUM ( < 1940) -0.4033

112-DUM (1940-1970) -0.4878

AG|I

Model I

C-~ fl-. I-value

1.9491 0.82

2.4276 1.68"

0.00002 2.21"

-0.0018 -0.25

0.00032 3.13"

0.00010 0.611

0.00003 0.20

- I .71"

-0.911

-I .42

Co.IT. t-value

3.71197 0.90

4.46467 1.72"

0.00003 : 1.97"

-0.10048 -0.39

0.0005 2.99"

0.0002 0.66

0.11001 0.42

54.993 - I .73-

-0.6556 -0.117

-0.1q662 , - I .39

ge,bL,

Ct~ff. l-value~

1.5239 0.65

2.3358 1,74"

0.00(102 2.53"

-0.0089 -I.07

0.00O32 3. I I "

0.00009 0.64

O.O0(X~ 0.56

-29.349 -1.67-

-I.4277 - I .49 -

-0.0259 1.52-

1611 No. OBS 168 166

Adj. R 2 . . . . . . . . . . . .

X ~ 27.q~ 27,95 28.24 27,76 (r 0 (9dl) (gdf) (qdl)

M odel I| !~,gi,

C(wI[. C-values

2.91107 I 0.72

4.6091 1.711"

0.00003 2.24"

-0.0160 1.13

0.0003 2.9T

0.0002 0.65

O.O~X}2 0.72

-53.816 - 1.69-

-2.1512 -1.30-

0.0398 1.37-

168

It should be noted that the t-statistics are based on asymptotic distribution theory for probit and logit. As such we do not make comparisons across estimators. The X2(df) is the chi-square statistic for the overall association with maximum likelihood methods. *These parameter estimates are significant at the 95 percent level of confidence. **These parameter estimates are significant at the 90 percent level of confidence.

This finding suggests that black households in this sample took advan- tage of the residential energy conservation tax credit program. The vari- able has the a priori expected sign. The larger the credit, the lower the after tax price and the stronger the impulse to make an energy-conservation improvement.

One distinguishing feature of the results in table 1 is a test of the interaction effect of household age and age of housing structure (AGH). 17 This factor may reflect any one of a number of considerations including the inability of older black households to increase their residential loca- tion choices because of housing market discrimination. We find that age has a negative influence on energy conservation improvement for blacks living in older buildings (coefficient of AGH is negative). It should be

Page 10: The effect of energy conservation tax credits on minority household housing improvements

132 The Review of Black Political Economy/Spring 1996

noted that the results are quite comparable over estimator in terms of the statistical significance of the factors postulated to influence energy con- servation improvements.

Next, we examine the results for Hispanics. The estimates for equation (6) is given in table 2. The results show that the expectation of higher average energy prices (PE) are an important determinant of energy con- servation choice for Hispanic households. The results also show that higher income Hispanics measured by INC are more likely to make insulation improvements. The number of heating degree days (HDD) is also a statistically significant determinant of insulation choice.

The structural characteristics of the house (AREA), H1-DUM and H2-DUM are generally statistically insignificant. The implication here is that structural effects are not important to Hispanics for investing in insulation equipment. The variable measuring the age of the household head is shown to effect energy conservation choice in the predicted di- rection, although it does not differ statistically from zero. The net-tax price measure of energy conservation capital has the hypothesized nega- tive sign and significant at the ten percent level. This finding implies that tax credits provide a substantial stimulus to Hispanics for installing en- ergy conservation equipment. The coefficient for the age-housing struc- ture interaction variable (AGH) is positively signed, and statistically in- significant. 18 Finally it should be noted that the results are quite compa- rable over estimator in terms of the statistical significance of the factors postulated to influence energy-conservation improvements.

IMPLICATIONS AND CONCLUSIONS

The present analysis extended earlier energy-conservation improve- ment choice findings by examining the effects of socioeconomic mea- sures, an array of housing structure characteristics, external temperature conditions, and the role of the residential energy tax credit within two model specifications and two estimating methods, for a sample of black households and Hispanic households respectively. In terms of the statisti- cal significance of the variables of interest, we find strong and consistent support that higher income blacks and higher income Hispanics are more likely to invest in energy-saving conservation equipment. Also, while older blacks are less likely to invest, older Hispanics are not. Blacks consider the size conditions of the house; Hispanics do not. Blacks and Hispanics also respond differently to changes in external temperature conditions. While blacks are more responsive to changes in cooling de-

Page 11: The effect of energy conservation tax credits on minority household housing improvements

Williams and Poyer 133

gree days, Hispanics are stimulated by changes in heating degree days. Of particular interest to this study, we find that tax credits for energy conservation expenditures have an important effect on insulation activity for both blacks and Hispanics. Our support for the importance of these factors is reinforced across the different estimators and respective esti- mating equations.

Lastly, this research does suggest that rather than accepting empirical results from a single study for energy-conservation choice for applicabil- ity to all types of households, further research should be directed to identify the ways in which particular variables could influence behavior for households disaggregated by socioeconomic type.

A C K N O W L E D G M E N T

Support for this research was received from the Department of Energy, Office of Minority Economic Impact. Views expressed in this article are those of the authors and do not necessarily reflect the view or opinion of either the Department of Energy or Argonne National Laboratory. We are indebted to an anonymous referee for some thoughtful comments and insights on an earlier draft of this article. All errors remain the responsibility of the authors.

N O T E S

1. M. J. Walsh, "Energy Tax Credits and Home Improvements," Energy Eco- nomics (1989): 275-284.

2. Pitts and Wittenbach, 1981. 3. See Walsh, op. cit. 4. See Walsh, op. cit. 5. See Pitts and Wittenbach, op. cit. 6. G. Garofalo and R. Molhatra, "Manufacturing Capacity Utilization Estimates

by Region," paper presented at the Western Economic Association Meetings, Se- attle, Washington, 1991.

7. Walsh, op. cit. 8. Energy Information Administration, Residential Energy Consumption Survey:

Housing Characteristics 1987, U.S. Department of Energy, Washington, D.C. 9. Hispanics in the sample are heavily concentrated in Texas, New Mexico, Ari-

zona, California, Chicago, Miami and New York. Blacks in the sample are more evenly distributed among regions.

10. We are indebted to Garofalo and Molhatra for making these data available to US.

11. B. C. Field and C. Grebenstein, "Capital-Energy Substitution in U.S. Manu- facturing," Review of Economics and Statistics (1988): 207-212.

12. T. A. Cameron, "A Nested Logit Model of Energy Conservation Activity by Owners of Existing Single Family Dwellings," Review of Economics and Statistics (1985): 139-149.

13. T. D. Wallace, "Pretest Estimation in Regression: A Survey," American Jour- nal of Agricultural Economics (1977): 57-82.

Page 12: The effect of energy conservation tax credits on minority household housing improvements

134 The Review of Black Political Economy/Spring 1996

14. A. S. Golberger, Econometric Theory, (1964), 249-259, New York: Wiley. 15. M. Gunderson, "Retention of Trainees: A Study with Dichotomout Depen-

dent Variables," Journal of Econometrics (1974): 79-93. 16. We tried other interaction terms in several alternative specifications of the

model. For example, we examined the interaction between income (INC) and heat- ing degree days (HDD); income (INC) and cooling degree days (CDD); and income (INC) and the net-price of energy saving capital (NTP). We would expect those households that are more informed of possible cost-savings from insulating would insulate at a lower number of heating and cooling degree days. In fact, the income variable may be capturing an awareness affect. Interacting INC with HDD and CDD with INC would allow for differing reactions to temperature variations among those householders who are more or less aware of the potential cost-savings. Lastly, the interaction between INC and NPT would allow for different price elasticities of demand between high-income and low-income black households. None of the coeffi- cients of these interaction terms were statistically significant.

17. We also included an aged-squared term in the model. The aged-squared nonlinearity term is intended to test whether the youngest households are more likely to move and therefore are less likely to invest in energy-saving equipment, because unlike the more elderly households, they do not expect to have a long enough tenure to reap the benefits of their energy-saving investments. Alternatively, testing the aged-squared term could highlight whether the young respond differently than the older householders because they expect their insulation investments to be capitalized in the value of their homes and will be recovered in the sale of the house. This coefficient was not statistically significant.

18. We considered the interaction between HDD and INC; CDD and INC; and NPT and INC in alternative specifications of the model for Hispanics (see endnote 3 for a discussion of these variables as they relate to blacks). The interaction terms in the specifications for Hispanics show that HDD and INC, and CDD and INC are both statistically significant. One interpretation is that higher income Hispanics are more informed of the potential cost-savings from insulating and tend to make the insulation improvements to reap these benefits. However, we find that coefficient of the interaction between NTP and INC is statistically insignificant.