Journal
of Public
Economics
EFFECTS
36 (1988) 259-267.
OF
TAXES
ON
(North-Holland)
NON-MARKET
WORK
The Swedish Case Lennart University
Received
of Gothenburg,
March
FLOOD* 411 25, Gothenburg,
1987, revised version
received
Sweden
May 1988
It is often assumed that high marginal taxes imply less work on the market and more work outside the market. A large amount of research has been done studying the effects of taxes on labor supply, but the effects on work outside the market are hardly known. Here we use information from a Swedish survey about households market and non-market activities and estimate the effects of marginal taxes as well as other socioeconomic variables on the amount of repair and maintenance activities performed. According to our results there is no positive effect of the marginal tax rate on home production. For females, we even find a significant negative correlation between repair work and marginal taxes.
1. Introduction This paper considers the statement that high marginal taxes can have an effect on the time spent on work outside the market. Do high marginal taxes imply more time spent on repair and maintenance of your home or other durables such as cars, boats, etc.? We will try to answer this question using a data base called HUS (Household Market and Non-market activities) [Klevmarken and Olovsson (1986)]. This data base includes rich information of a representative sample of individuals in Swedish households in 1984. Each individual was asked how many times they have done repair and maintenance activities during the last two weeks. Here we excluded all respondents who were over 65 years of age. The resulting frequencies for 998 males and 1,048 females are presented in table 1. About 38 percent of males did not report any repair activities, which should be compared to 57% for females. The mean value is 2.2 for men and 1.2 for women. Thus, as expected, these activities are clearly male dominated. It should, however, be emphasized that it is the number of times that repair and maintenance activities were done that has been measured. The length or intensity each time is not known. In what follows we derive an estimable relation explaining the frequency of repair and maintenance activities by a set of exogenous variables, including *I have benefited 004772727/88/$3.50
from comments c
by Anders
Klevmarken
1988, Elsevier Science Publishers
and two anonymous
B.V. (North-Holland)
referees.
260
L. Flood, Effects
of taxes
Table Frequency
of repair
on non-market
1
and maintenance
Males Times 0
1 2 3 4 5 6 I 8 9 10 11 12 13 14 15 17 18 23
work
activities.
Females Frequency
Percent
Frequency
Percent
379 156 131 96 12 44 38 24 18 16 8 3
38.0 15.6 13.1 9.6 7.2 4.4 3.8 2.4 1.8 1.6 0.8 0.3 0.3 0.3 0.3 0.2 0.1 _
599 164 112 59 38 21 12 16 9 6 4 2 1 2 2
51.2 15.6 10.7 5.6 3.6 2.0 1.1 1.5 0.9 0.6 0.4 0.2 0.1 0.2 0.2 _
3 2
1 1
_ 0.1 _
0.1
the marginal tax rate. The variables used are presented in table 2 and the resulting estimates in table 3. The paper concludes with an interpretation of these estimates.
2. The model’ It is assumed that individual welfare goods and services (X) and leisure (L).
(U) is a well-behaved
function
u = U(X, L). Total consumption goods (X,),
of
(1) is composed
of market
goods (X,)
and home
x=x,+x,.
produced
(2)
It is assumed that welfare is not affected by the composition of X, that is su/zx, = C?U/dX,. Home produced goods are produced by time spent on repair and maintenance activities (H): ‘For Gronau
a more detailed discussion of the model does not consider the effects of taxes.
used here see Gronau
(1977,
1980). However,
X,=f(H), The maximization (1) The budget
L. Flood, Effects of taxes on non-market work
261
f’>O,f”
(3)
of welfare is subject contraints,
X,=wN+
to two constraints.
V-t(l),
(4)
where w is gross wage per hour, N is hours of market work, I/ is non-labor income and t is the total amount of income taxes paid, which is a function of taxable income 1. Taxable income is defined as wN + V-D, where D is total deductions. (2) The time constraint, N+H+L=T The tax function
(5) in (4) is approximated
by
(6)
t = tj + t;.(Z- Z,),
where lj is the smallest taxable income in the observed tax bracket j, tj is the tax payable at that income and t; is the marginal tax rate. Maximization of welfare subject to the constraints yields the equilibria: MRS=f’=w’,
(7)
where MRS is the marginal rate of substitution between leisure and consumption, f’ is the marginal productivity in repair and maintenance and w’ is the maginal after tax wage, that is w’= w( 1 -t’). Following Gronau (1980) a particular functional form for f’ is assumed:
f:=-crH,+Bo+
~ BjZij,
i=l ,...,n,
(8)
j=l
where subscript i denotes individuals and Zi is a vector of exogenous variables. It is assumed that the individual decides how much time to spend on repair activities by comparing f’ and w’. If w’ >f’, then no repair work is undertaken, and if w’=f’, then he will spend some positive amount of time on this activity. This decision rule can be written: Hi(w;-f;)
=O.
(9)
262
L. Flood,
Effects of taxes on non-market
Eqs. (8) and (9) detine the following
Hi=fl,* + ~ PlZij+a*w;,
equation
if
work
for repair work:
Hi>O,else H,=O,
(10)
j=l
where @T= p,Ja, j = 0,. ..,k, a*= - l/a. Eq. (10) corresponds to a Tobit model [Tobin (1958)]. However, since Hi in our case is a discrete variable, i.e. the number of times repairs and maintenance are done, a Poisson specification is more appropriate. The Poisson model is designed for count data. Recent applications can be found in Hausman, Hall and Griliches (1984) Robin (1987) and Allessie, Gradus and Melenberg (1987). The Poisson specification has several advantages. There is no need to use truncated distributions, as is the case in the Tobit model, since the zeros are a natural outcome of the Poisson specification. The integer property of our data is handled directly and the estimation is simple. It is assumed that the reported number of times of repair activities have independent Poisson distributions with parameters &. Hence,
P(Hi= Y)= exp ( - AJA;/r!, where Y is the value assumed
lOg~i=/3X+
~
j=l
by
(11)
Hi,and
pj*Zij+Cr*Wl.
(12)
The exogenous variable that we focus our interest on is the marginal tax rate. The definition of this variable is documented in Klevmarken and Olovsson (1986). Here it suffice to say that we take account of the tax rules and use taxable income for the income year 1984, which has been obtained mostly from register data but also from the survey. The effects of housing allowance and cost of childcare have not been considered in the calculations of the marginal rates. The marginal tax rate does not enter (12) directly as an exogenous variable. It is the marginal net wage, w’, that is important for the individual’s decision to spend time on the repair and maintenance activity. The effects of the marginal taxes can be estimated indirectly since w’= w(1 -t’). Thus, the reason for not including the marginal tax rates explicitly as an exogenous variable is that we assume that the decision to allocate time on non-market work is based on the marginal net wage. However, marginal taxes also affect this decision since they affect the marginal net wage. In principle, the marginal tax rate not only affects labor income but also non-labor income. However, it is difficult to incorporate non-labor income taxes in the analysis
L. Flood, Effects of taxes on non-market work
263
since only a part of the non-labor income is taxable. Thus, the marginal tax effects that we study here are only the labor income tax effects. Note the difference between our model and the traditional labor supply model. The inclusion of the marginal net wage as an exogenous variable in the labor supply model, results in a problem of endogeneity. This problem is due to the fact that the marginal net wage is a function of marginal taxes, which in turn is a function of the endogenous variable, hours of work. Several methods have been suggested in order to solve for this problem; see, for example, Hausman (1979). However, in our model this is not a problem since the endogenous variable is the frequency of repair work. In order to study the isolated effects of the tax rate, other relevant variables explaining the frequency of repair and maintenance must be included. Following Hill and Juster (1985) we regard the decision to allocate time to non-market work as being determined by three sets of factors: Constraints representing short- and long-run commitments; productivity as reflected by experience or skill to perform various types of repair and maintenance activities; and preferences for the activities themselves. Constraints are represented by variables such as time spent on market work, ownership of house, ownership of leisure house or boat, number of cars in the household, number of children below seven years of age, number of adults in the household, and household non-labor income in thousands of kronor. Market work is considered as a constraint since more time spent working means less time to spend on other activities. Market work is measured as a share, namely working hours per week divided by total available time a week. Females spend about 26 percent of their time on market work compared to 37 percent for males. Note that commuting is included in market work. Ownership of a house is a dummy, where one indicates house, otherwise zero. Ownership of leisure house or boat is defined in the same way. Productivity is measured by age and number of years in school. The square of age is also included, since we expect the effect of age to be nonlinear. Wage per hour before tax has also been included since it is assumed that wage rates by themselves have negative effects on home production. It is therefore important to standardize for wage rates. Finally, we also include a measure of preferences. This measure has been obtained by asking the individuals how much they enjoy doing repair and maintenance activities. Zero indicates that they did not enjoy it at all and ten that they enjoy it very much. These questions were designed to ensure that we measured satisfaction with the activity itself rather than satisfaction with the outcome of the activity. We adjusted the resulting &lo scores to eliminate systematic differences across individuals due to what is considered as a mean value on the &lo preference scale. This adjustment could be done since we asked about preferences for several different activities apart from
264
L. Flood, Effects of taxes on non-market work Table 2 Descriptive
statistics
of variables.
Males
Females
Variable
Mean
Std.
Min.
Max.
Mean
Std.
Min.
Max.
Repair activities Houseowner Leisure house Boat Number of cars Nllmber of adults Number of children Non-labor income
2.20 0.72 0.24 0.29 1.19 2.18 0.32 11.43 42.08 0.37 10.96 0.53 23.33 49.77 0.25
2.82 0.45 0.43 0.45 0.75 0.68 0.63 20.03 12.11 0.15 3.60 0.13 10.24 24.80 2.14
0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 19.00 0.00 1.00 0.00 2.91 6.38 -6.88
23.00 1.00 1.00 1.00 9.00 5.00 3.00 110.64 65.00 0.87 28.00 0.84 164.65 426.00 7.17
1.22 0.69 0.23 0.27 1.10 2.15 0.30 13.99 41.42 0.26 10.63 0.42 24.27 41.61 - 1.09
2.15 0.46 0.42 0.44 0.72 0.66 0.60 24.72 12.66 0.15 3.25 0.17 10.33 19.61 2.57
0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 18.00 0.00 3.00 0.00 4.34 7.58 -7.86
18.00 1.00 1.00 1.00 7.00 5.00 3.00 292.00 65.00 0.85 27.00 0.77 207.38 448.00 6.00
Age Market work Education Marginal tax rate Marginal wage rate Wage before tax Preferences
repair and maintenance. The mean value for each person across all activities was calculated, and then this mean value was subtracted from the response given for the repair activity. From table 2 it follows that males prefer the repair activity a bit more than the mean value of preferences, but for females it is the other way around. The descriptive statistics of the variables used are presented in table 2. From table 2 it follows that the average marginal tax rate for males is about 50 percent and for females about 40 percent, the highest rates are about 80 percent. Using the variables discussed above, model (11) has been estimated, for men and women separately, by the maximum likelihood method. The resulting estimates are presented in table 3.
3. Results Columns (1), (2), (4) and (5) of table 3 display the estimates and corresponding t-valus for males and females, respectively. In columns (3) and (6) we also present the results in the form of elasticities. As expected there are, in general, positive and significant effects of variables expressing commitments such as ownership of house, leisure house, and the number of cars. The effects of these variables are in general stronger for females. Ownership of a boat has no significant effect. We expected a negative effect of the number of adults in the household since they represent a resource that decreases the workload. This negative effect can also be found for females, but for males there is no significant effect. The effect of children is positive but non-significant. Since many of the
L. Flood, Effects of taxes on non-market work
265
Table 3 Estimated
coefficients
and t-values. Females
Males Variable
Estimate
t
Constant Houseowner Leisure house Boat Number of cars Number of adults Number of children Non-labor income Age Age/square 1,000 Market work Education Marginal wage rate Wage rate before tax Preferences
-0.5523 0.4599 0.0753 0.0004 0.0618 0.0126 0.0272 0.0020 0.0057 -0.1192 0.9582 0.0133 0.0032 0.0034 0.1099
1.85 1.45 1.44 0.01 2.15 0.36 0.73 1.69 0.40 0.73 5.60
1.94 1.17 3.08 9.84
Elasticity
_ 0.33 0.02 0.00 0.07 0.03 0.01 0.23 -0.18 0.35 0.15 0.07 0.17 0.03
Estimate
t
- 1.9806 0.5004 0.3242 - 0.0638 0.1556 -0.1166 0.0814 0.0013 0.0488 - 0.4078 -0.0372 0.0383 0.0193 - 0.0079 0.0676
5.07 6.55 4.83 0.97 3.61 2.33 1.53 1.18 2.69 1.97 0.15 3.98 3.16 2.55 6.03
Elasticity
_ 0.35 0.07 - 0.02 0.17 -0.25 0.02 0.18 0.63 _ -0.01 0.41 0.47 -0.33 -0.07
repair and maintenance activities can be done together with children this results is not implausible. Non-labor income has a positive effect. This might be unexpected since higher income means that market services or laborsaving equipment could be bought, which should reduce time on repair acitivites. However, higher income also implies a larger house/leisure house or a higher number of cars which should increase the time on repair and maintenance activities. The effect of market work is not the same for men and women. An increase in market work means a higher frequency of repair activities for males. An increase in market work by 1 percent implies an increase in repair activities by 0.35 percent. At first sight the effect might be expected to be negative since more market work means less time to spend on non-market activities. However, the positive effect on males can be explained by differences in productivity. Men with higher productivity work more both in and outside the market. The skill acquired in market work might be important also for the work done on repair and maintenance activities. Thus, the amount of market work is also a measure of productivity as well as age and education. Market work has no significant effect on females. The effect of education is positive, as expected, for both sexes. Age has a positive coefficient and age-square a negative, but these effects are not significant for males. It was assumed that wage before tax would have a negative effect on repair activities; however, this is only true for females. According to the results, a 1 percent increase in the gross wage decreases female repair activities by 0.33 percent and increases male activities by 0.17 percent.
266
L. Flood, Effects of taxes on non-market work
As expected, the effect of preferences are positive; note the high t-values of these estimates, Thus, preferences are indeed relevant in explaining the allocation of time. However, with the exception of Hill and Juster (1985), preferences have not, at least to our knowledge, been used before as an explanatory variable in time allocation models. From the discussion above it follows that the results seem quite sensible and we proceed to discuss the effects of marginal taxes, Marginal wages have a significant effect on the amount of repair activities done by females; the effect on males is however insignificant. From the point estimates in table 3 it follows that an increase in the marginal wage rate implies more repair activities by both sexes, but the elasticities suggests a much stronger effect on females. Since w’ = w( 1 -t’) this means that an increase in the marginal tax rate implies less repair work. In order to get a clear picture of the size of the estimated effects, we present tax elasticities, i.e. the effect on the amount of repair activities by a small change in the marginal tax rate. The elasticities have been calculated as c(*tii’, where W and ?’ are the gross wage rate and marginal tax rate evaluated at their mean values. The resulting male elasticity is -0.08 and the corresponding female elasticity is -0.34. From these elasticities alone it is not obvious what the total effect will be on the time spent on repair and maintenance, apart from the fact that the time will decrease due to an increase in the marginal tax rate, which of course in itself is an interesting result. However, using additional information from the HUS survey it is possible to get a rough estimate of the effects on the total time spent doing repair work. From a time-use study, which is a part of HUS, it follows that the mean time spent on repair and maintenance activities each day is about three hours for men and about two hours for women, for those who actually perform these activities. The frequencies are also known. On average the participating male and female reports 2 and 1.5 activities per day, respectively. Thus, the mean time spent on each activity is one and a half hours for males and about one hour for females. If we assume that this is the mean time spent on each reported repair activity, the resulting time-use per week is about three hours for males and about one and a half hours for females. Thus, a 1 percent increase in the marginal tax rate decrease the time spent doing repair work by about 15 minutes for males and about 30 minutes for females per week. Thus, our results indicate, contrary to what is often assumed, a negative correlation between repair work and marginal tax rates. However, this effect is only significant for females. A 1 percent increase in the marginal tax rate implies on average that females spend about half an hour less time on repair and maintenance activities each week. However, it should be stressed that the repair activity is only a small part of all home production activities, especially for females. Thus, our results do
L. Flood, Effects of taxes on non-market work
261
not rule out the possibility of a positive correlation between marginal tax rates and time spent on home production, as a whole. But, nonetheless, it is still an interesting result that marginal taxes should be negatively correlated with repair work. One possible explanation is that females with a low marginal tax rate are working less in the market and therefore have more time to spend on repair work. However, this conclusion does not seem to be confirmed by our results. The amount of market work was included as a regresssor in order to control for this, and no significant effect of market work on repair activities can be found. This result must perhaps instead be viewed in a household context. In households where the husband is a highincome earner, the female can spend less time on market work and therefore her marginal tax rate is low. However, she might spend a lot of time doing repair and maintenance work since the household has a high income and therefore has a greater likelihood of owning a big house/leisure house as well as other time-consuming durables. The discussion above suggests an interesting development of the model used here. As a future research work it would be interesting to study the simultaneous allocation of time between market and non-market work as well as between spouses.
References Allessie, R., R. Gradus and B. Melenberg, 1987, The problem of not observing small expenditures in a consumer expenditure survey, Research Memorandum, Department of Economics, Tilburg University. Gronau, R., 1977, Leisure, home production and work: The theory of the allocation of time revisited, Journal of Political Economy 85, no. 6. Gronau, R., 1980, Home production a forgotten industry, The Review of Economics and Statistics 62. Hausman, J., 1979, The econometrics of labor supply on convex budget sets, Economics Letters, no. 3. Hausman, J., H. Hall and Z. Grihches, 1984, Econometric models for count data with an application to the patents-R&D relationship, Econometrica 52, no. 4. Hill, MS. and F.T. Juster, 1985, Constraints and complementarities in time use, in: F.T. Juster and F.P. Stafford, eds., Time, goods, and well-being (Survey Research Center, Institute for Social Research, University of Michigan). Klevmarken, N.A. and P. Olovsson, 1986, HushalIens ekonomiska levnadsforhallanden (HUS), Teknisk beskrivning och kodbok (Department of Economics, University of Gothenburg). Robin, J.M., 1987, A new model to estimate Engel curves from budget expenditure data, Working paper (INRA, Paris). Tobin, J., 1958, Estimation of relationships for limited dependent variables, Econometrica 26.