Labor demand, labor costs, health and disability

Transcription

1 Labor demand, labor costs, health and disability Anna A. Godøy and Ingrid M. S. Huitfeldt October 10, 2015 Abstract We use exogenous variation in local payroll tax rates to estimate effects of higher labor costs on earnings, employment and program participation of workers with poor ex ante health endowments, as measured by low birth weight. From 2003 to 2006, almost half of Norwegian municipalities experienced forced harmonization of regional payroll taxes following a ruling in the European courts system. Exploiting the resulting variation in tax rates, we show that individuals with poor health endowments were disproportionately affected by the tax hike, experiencing larger earnings losses, lower labor force participation and higher rates of disability insurance uptake. Our findings indicate that high labor costs may be one factor keeping workers with poor health out of the labor market. Keywords: Labor demand, labor force participation, disability insurance JEL classification: H25, I14, J21, J23 This research is part of the project Striving for excellence, learning to cope, financed by the Norwegian Research Council (grant ). Data made available by Statistics Norway have been essential. Institute for Social Research, P. box 3233 Elisenberg, 0208 Oslo, Norway. a.a.godoy@socialresearch.no Ragnar Frisch Centre for Economic Research, Gaustadalléen 21, 0349 Oslo, Norway. i.m.s.huitfeldt@frisch.uio.no 1

2 1 Introduction The rise of disability insurance (DI) rolls in many western countries has attracted substantial attention from economists and policymakers (see, e.g., Von Wachter et al. 2011, Kostøl & Mogstad 2014). Much of the policy debates and research on this topic is focused on factors that influence labor supply, e.g. the eligibility requirements and financial incentives of DI programs and replacement rates, and activation requirements on DI enrollment. In the present paper, we focus on factors influencing firms demands for workers who may be on the margins of the labor market. Specifically, we examine whether an increase in payroll taxes has a differential effect on the labor market outcomes of workers with poor health endowments, as measured by birth weight. To identify the impacts of payroll tax hikes, we exploit a forced change in the system of regionally differentiated payroll tax rates following a ruling in the European courts system. Since 1975, Norway has had a system of geographically differentiated payroll tax rates, intended to encourage business in remote and rural areas. In 2002, this policy was ruled to be in violation of trade agreements, and as a result, payroll tax rates were reduced over a three year period from 2003 to Crucially, this change was not made in response to regional development trends, but rather imposed as a consequence of European trade agreements in spite of the wishes of local policymakers. This policy shift thus created exogenous variation in tax rates across regions and time. From economic theory, the wage and employment effects of changes in payroll tax rates are ambiguous, as they depend on a number of factors, notably the labor supply elasticity and the wage setting scheme. In the absence of minimum wage laws or other wage rigidities, payroll taxes will be shifted fully or partially onto workers wages, to a degree that depends on the elasticities of labor demand and labor supply with respect to wages. In particular, if labor supply is inelastic, the entire tax burden will be shifted onto wages, and employment will remain unaffected. However, the centralized nature of collective wage bargaining in Norway implies that there will be wage rigidities, at least in the short term. In that case, taxes will only partially be shifted onto wages, with firms reducing their labor demand in response to higher labor costs, even when labor supply is completely inelastic. There are several potential reasons to expect workers with poor health endowments to be especially affected by such a drop in labor demand. Low birth weight is associated with lower adult IQ and educational attainment(black et al. 2007). In addition, low birth weight is associated with a number of health conditions, so this group may, on average, have more frequent periods of sick leave, at an additional cost to employers. Given the system of collective wage bargaining, these productivity differences may not be fully reflected in wages. Drawing on the search and matching literature (Pries & Rogerson 2005), when faced with a tax hike in combination with a (de facto) binding minimum wage, firms require a higher minimum expected match productivity when hiring. Moreover, in the presence of sorting, low birth weight persons may be overrepresented in low productivity firms, who may be more vulnerable to tax hikes. Poor health 2

3 may reduce geographic mobility, making it harder for this group to respond to reduced local labor demand by moving to a better labor market. Finally, to the extent that at least some tax shifting occurs, and this is interpreted as lower expected future earnings, this may increase the relative appeal of DI receipt. A number of papers have investigated the incidence of payroll taxes issue empirically. Saez et al. (2012) use data from a Greek reform affecting payroll tax rates in younger cohorts to identify long run effects of payroll tax rates. Using a regression discontinuity design approach, they find that higher payroll tax rates did not affect wages or labor supply. Bennmarker et al. (2009) use a reduction in regional payroll tax rates in northern Sweden to analyze effects of payroll tax rates, and find some evidence of employment effects, driven by changes in entry and exits of firms. An important difference between their paper and ours is that in our case, the changes in regional payroll tax rates were imposed from outside, rather than decided by local policymakers. Dale-Olsen (2014) uses the same payroll tax reform as the present paper to analyze wage and employment effects among private sector workers. Using a differences-in-differences approach, he finds that in the short run, some wage shifting occurs. While his paper finds little evidence of short term labor supply effects, he finds increased DI participation rates in the long run. The present paper is also related to the literature linking unemployment and disability (Duggan et al. 2003, Rege et al. 2009, see, e.g.), that find that disability insurance uptake increases substantially after job loss. These effects are found to be driven by a combination of worsened health and drop in expected future earnings. A number of papers have examined the effects of low birth weight on adult outcomes (see, e.g., Black et al. (2007), Royer (2009)). Low birth weight is found to lead to worse education outcomes, lower income, and higher rates of DI receipt. There is less research done on the importance of birth weight in explaining the differential impact of economic shocks. One exception is Bharadwaj et al. (2013), who examine the impact of birth weight among adults with strong labor market attachment during the course of the Swedish recession of the 1990s. They find a differential impact: adults who had poor infant health had higher rates of job loss and unemployment insurance takeup during the recession. Moreover, their results indicated that this was not driven by intermediating factors like educational attainment or occupational choice. The rest of the paper is organized as follows: Section 2 gives a brief review of relevant institutions, presenting the reform in payroll taxes, before presenting the sample. Our empirical models and findings are presented in section 3. Section 4 concludes the paper. 2 Institutions and data In this paper, we study younger workers (age 20-39), observed during the first half of their labor market careers. This section gives a brief review of the most relevant institutional context: the educational system, the wage setting process 3

4 and some relevant social insurance arrangements. Next, we present the system of differentiated payroll tax rates, and the imposed tax reform following the 2002 EFTA court ruling. Finally, the sample selection criteria are presented together with some summary statistics of the data. Institutions Wage setting in Norway is characterized by an important presence of collective bargaining, leading to a compressed wage structure (Barth et al. 2014). The process can be described as a two-tier system, starting with centralized bargaining carried out by national labor unions and employers associations. After this, local wage negotiations take place between employers and firm level unions. Unemployment insurance is universal, with eligibility determined by earnings history; the replacement rate is around 66%. People who are out of work and who do not qualify for unemployment insurance can apply for social assistance, provided they have no other way of supporting themselves. Social assistance is means tested against household income and assets, and is intended to cover basic subsistence costs only. Employees who are unable to work due to an illness or injury are entitled to paid sick leave for up to 1 year, at 100% replacement rate. After this, they are eligible for temporary disability insurance, at substantially lower replacement rates (around 66%). Individuals whose illness or injury leads to permanent inability to work, can apply for disability insurance (DI). Persons with little or no labor market experience are entitled to a guaranteed minimum level of (temporary or permanent) disability insurance payments 1. While education (secondary school and university) is provided by both public and private institutions, tuition-free public schools and universities have far higher enrollment figures compared to private institutions. University students are eligible for financial aid for up to eight years of higher education, conditional on passing exams. Enrollment and graduation rates are generally found to be countercyclical (Reiling & Strøm 2015). Payroll taxes Payroll tax rates in Norway are geographically differentiated according to the location of the firms workers 2. Geographically differentiated payroll tax rates were introduced in 1975, in an effort to subsidize employment in struggling rural and remote areas. The country was divided in 5 tax zones, with payroll tax rates ranging from 14.1% in the central areas to 0% in the northernmost regions. The exact delimitation of the payroll tax zones have been changed repeatedly from 1975 to the present day, typically in response to changing economic con- 1 Currently, the guaranteed minimum permanent disability payment for a single person is 223,369 NOK, equivalent to 24,364 EUR with current exchange rates. 2 This was changed to the location of the plant from 2007, i.e. after the period studied in this paper. 4

5 Figure 1: Payroll tax zones Payroll tax rates Note: Figure shows pre-reform payroll rates. ditions. The payroll tax zones effective during the period studied in this paper are illustrated in figure 1. In 2002, the EFTA Court issued a ruling stating that the system of geographically differentiated payroll tax constituted an illegal requiring payroll tax rate harmonization. The precise implementation of this ruling was subject to negotiations between EFTA officials and the Norwegian government, that lasted through the spring/early summer of In the end, it was decided that zone 5 would be exempt from payroll tax harmonization. Harmonization would happen immediately in zone 2 (where the initial payroll tax rate was 10.4%). In zones 3 and 4, where the initial tax rates were significantly lower, harmonization would happen gradually over a period of 3 years. This harmonization schedule was followed until the end of In 2006, European rules on regional support policies were changed in a way that implied that regionally differentiated payroll taxes would no longer be in conflict with EU competition laws. As a consequence, geographically differentiated payroll tax rates were re-introduced January 1st Figure 2 illustrate the development of payroll tax rates in the different regions. The changes in payroll tax rates in the years are of particular interest as they were imposed by a European court, and not by local authorities. In other words, the change was not motivated by changing economic conditions in the affected regions. This allows us to overcome the problem of policy endogeneity. Moreover, as can be seen from the map in figure 1, the changes covered a significant part of the country: 230 of 431 Norwegian municipalities experienced changes in the payroll tax rate as a result of the EFTA court s ruling. In a response to this unwanted tax change, additional funds were established 5

6 Figure 2: Payroll tax rates Payroll tax Rate Year Sone 1 Sone 2 Sone 3 Sone 4 Sone 5 Note: Figure shows payroll rates from 2001 to 2008 in the five payroll tax zones that existed pre-reform. After 2007, some changes were made to increase PRT in larger towns in zone 4, as well as other minor changes in the payroll tax system. to support regional development projects. In addition, it was decided that the affected municipalities would receive additional transfers to compensate for the higher labor costs of municipal workers. For this reason, the reform might be expected to have the largest effects on private sector workers (as well as job seekers applying primarily to private sector jobs). A central point in our analysis is whether the change in payroll tax starting January 1st and the reversal in were anticipated. While this is not directly observable in our data, we can get an impression by looking at media coverage on payroll taxes between January 2002 and December Media coverage of the payroll tax rates peaked in in particular when focusing on stories that mentioned the EFTA court. However, it is not clear that firms were able to immediately respond to these discussions, especially given the high level of uncertainty about the exact nature of tax changes. In the main analyses, we therefore include 2003 in the estimations. Sample The sample includes all persons born between 1967 and People are included in the sample from the year they turn 20. Individuals are locked to their home municipality January 1st 2003, in order to avoid problems of endogenous moving (in alternative specifications, we allow for people to move across municipalities). The sample is then merged to data on annual labor earnings, and 6

7 participation in various programs (education, social welfare, unemployment and disability). Data on birth weight is available for individuals born in 1967 onwards. Our main measure of health status is low birth weight (defined as birth weight < 2500 gram), a threshold that is widely used in the literature. Requiring information on birth weight implies that all foreign born individuals are excluded from the sample. Other than this, we retain a large share of each cohort. The resulting sample is an unbalanced panel following 1,068,281 individuals, of which 223,038 are residing in the affected municipalities (retained in the main analysis sample). Table 1 presents summary statistics of the sample. Zone 1 is the largest, and also the most educated, with the highest average labor earnings. This region covers the Oslo area and most larger cities. The zones affected by the payroll tax harmonization, zones 2, 3 and 4, are comparatively more similar. Roughly 30% of the sample have only compulsory schooling, while 17-20% have at least a 3- year college degree. In the affected regions, between 3.5 and 3.9% of individuals have low birth weight. Table 1: Summary statistics Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 mean mean mean mean mean Payroll tax (2002) Age Unknown or elementary ed High school College Female Low birth weight Earnings > log(w) Labor earnings Observations Empirical findings We model employment outcomes of individual i, living in municipality j (in 2003), in year t y ijt =θ j + θ t + CY jt + x it β x + P RT ijt β P RT + LBW i β LBW + (P RT ijt LBW ijt )β P RT LBW + ε ijt (1) Where θ j, θ t are municipality and year fixed effects. x it is observable individual characteristics; in the baseline model, this is a vector of age dummies. LBW i is an indicator of low birth weight (< 2500 g), P RT is the (log of the) payroll tax. 7

8 Equation (1) is estimated on the subsample of individuals who are living in one of the three affected payroll tax zones 3. Table 2 shows estimated effects on employment outcomes. Column (1) shows estimated effects on linear earnings. A one percent increase in the payroll tax rate reduces earnings by 7,113 NOK. Moreover, the model finds an additional negative effect on earnings for individuals with low birthweight - an additional reduction in labor earnings by 9,286 NOK, or more than double compared to persons with higher birth weight. Table 2: Labor market outcomes (1) (2) (3) Earnings Log earn. Pos. earn. ln(prt) (-2.21) (-2.58) (1.04) LBW (0.18) (-1.00) (0.18) LBW ln(prt) (-2.94) (-1.13) (-2.43) Observations t statistics in parentheses p < 0.1, p < 0.05, p < 0.01 Note: Table shows selected coefficients from estimation on (1) on persons in the affected municipalities. Models include municipality fixed effects, year dummies, age (5-year dummies) and county level linear time trends. ln(prt) is the log of the payroll tax, LBW is an indicator of low birth weight (< 2500 grams). Standard errors clustered at the municipality level. Next, we want to examine whether this is happening through cuts in wages or hours (the intensive margin), or through a reduction in the proportion who are employed - defined as having any labor earnings. Column (2) shows estimated effects on log earnings, i.e. excluding persons with zero labor earnings from the estimation sample. As our data does not have accurate information on hours worked, this reflects impacts on hourly wages, hours worked, or a combination of the two. There is a significant and negative effect of payroll taxes on log wages: a one percent increase in the payroll tax translates to a 0.06% drop in labor earnings. For log earnings, there is no significant differential effect for individuals with low birth weight. Column (3) shows effects on employment, defined in the widest possible sense as having labor earnings > 0. For the main effect, we find no significant impacts on the payroll tax rate on employment. However, the interaction term P RT ijt LBW ijt is negative and significant, indicating that higher payroll taxes reduce employment for this group. To summarize, increased payroll taxes lower labor earnings, and the effect is approximately twice as large for persons with low birth weight. This additional 3 Locking individuals to 2003-residence, ignoring moves before and after

9 effect appears to operate primarily on the extensive margin, through a reduction in employment rates. Table 3 shows effects on program participation. The four modeled outcomes are indicators equal to one if the person at any time during the year is enrolled in education (secondary school or university), registered unemployed (fully or partially, including participation in training programs offered through the unemployment office), receiving permanent disability benefits or receiving social assistance payments. Note that the four outcomes are not mutually exclusive: for example, it is possible for a single individual to be employed at the beginning of the year, become unemployed in the spring, and enroll in organized education the fall semester. Our approach allows us to identify the short term effects on employment and program participation. However, due to the nature of the imposed tax change (gradual rollout over three years, followed by a near-complete reversal), it is difficult to measure more long term impacts. Higher payroll tax rates significantly increase the share that is enrolled in school, as well as the share that is registered unemployed. For these outcomes, there is no direct effect of low birth weight, and no interaction effect between P RT and low birth weight. For DI receipt, there is no effect for the population as a whole. There is, however, a positive effect for persons with low birth weight. While significant at the 10% level only, this is an indication that higher payroll taxes could increase permanent DI receipt for this group. Finally, the model finds no significant effects of higher payroll taxes on social assistance takeup. Table 3: Program participation (1) (2) (3) (4) In school Unempl. Disability Soc. ass. ln(prt) (3.01) (2.53) (0.37) (1.61) LBW (0.78) (0.10) (0.79) (0.06) LBW ln(prt) (-1.47) (0.66) (1.79) (1.47) Observations t statistics in parentheses p < 0.1, p < 0.05, p < 0.01 Note: Table shows selected coefficients from estimation on (1) on persons in the affected municipalities. Models include municipality fixed effects, year dummies, age (5-year dummies) and county level linear time trends. ln(prt) is the log of the payroll tax, LBW is an indicator of low birth weight (< 2500 grams). Standard errors clustered at the municipality level. Next, the models are estimated separately by gender. One motivation for doing this is that the changes in payroll tax rates disproportionately affected the private sector. As discussed above, local governments were compensated 9

10 through additional transfers to compensate for higher labor costs. As a result, we would expect effects to be stronger for persons who work, or who are applying for work in the private sector. As a substantial part of our sample has low labor market attachment, conditioning on sector of employment is potentially problematic. Rather, we do the analysis by gender (which is pre-determined). The labor market in Norway is characterized by substantial gender segregation, where women tend to work in the public sector and men in the private sector. For this reason, we might expect stronger impacts for men than for women. Another consideration is connected with endogenous fertility and effects of children on labor supply decisions. The sample population is aged between 20 and 39 years, an age where many people consider having (additional) children. Almost all women in Norway take 6-10 months parental leave after giving birth, some stay home considerably longer, and a significant fraction work part time while caring for small children. Men, on the other hand, tend to take much shorter periods of parental leave (2-3 months), and a large majority return to working full time. Table 4 shows estimated effects on labor outcomes by gender. For the population as a whole, payroll taxes have stronger effects for men than for women. Looking at the differential effect for individuals with low birth weight, estimates are more similar: men (women) with low birth weight lose an additional 8273 (7061) NOK in labor earnings on average following a one percent increase in the payroll tax rate. As before, there are no differential effects of low birth weight on log earnings. The differential effect on the probability of having positive labor earnings is, perhaps somewhat surprisingly, only significant for women. However, it should be noted that splitting the sample reduces the significance of the estimated effects, so the differential effects by birth weight are significant on the 10% level only. Table 5 presents estimated effects on program participation for men and women. Payroll taxes increase educational enrollment for both men and women, though the effect is more significant for men. Men also have an increase in UI takeup. The differential impacts on persons with low birth weight are also only significant for men. Unlike their peers with normal birth weights, they do not respond to higher PRT by increased educational enrollment 4. Rather, they respond by higher rates of disability and social assistance. 4 Conclusions In a 2002 ruling in the European EFTA court, the Norwegian system of regionally differentiated payroll taxes was ruled to be in violation of European trade agreements. This resulted in a forced harmonization in payroll taxes from 2003 to 2006, affecting 206 municipalities. Centralized collective wage bargaining meant that the tax hike was only partially shifted onto wages, resulting in increased labor costs in the affected regions. 4 For men with low birth weight, the negative interaction effect of LBW P RT roughly cancels out the positive main effect of P RT. 10

11 Table 4: Labor outcomes by gender (a) Men (1) (2) (3) Earnings Log earn. Pos. earn. ln(prt) (-2.50) (-2.50) (0.46) LBW (-0.19) (-1.22) (-0.17) LBW ln(prt) (-1.66) (-0.30) (-1.36) Observations t statistics in parentheses p < 0.1, p < 0.05, p < 0.01 (b) Women (1) (2) (3) Earnings Log earn. Pos. earn. ln(prt) (-0.53) (-1.73) (1.10) LBW (0.32) (-0.24) (0.39) LBW ln(prt) (-1.70) (-0.78) (-1.83) Observations t statistics in parentheses p < 0.1, p < 0.05, p < 0.01 Note: Table shows selected coefficients from estimation on (1) on persons in the affected municipalities, by gender. Models include municipality fixed effects, year dummies, age (5- year dummies) and county level linear time trends. ln(prt) is the log of the payroll tax, LBW is an indicator of low birth weight (< 2500 grams). Standard errors clustered at the municipality level. 11

12 Table 5: Program participation by gender (a) Men (1) (2) (3) (4) In school Unempl. Disability Soc. ass. ln(prt) (2.91) (3.02) (0.53) (1.54) LBW (1.61) (1.27) (-0.15) (-0.64) LBW ln(prt) (-1.89) (-0.79) (2.11) (1.65) Observations t statistics in parentheses p < 0.1, p < 0.05, p < 0.01 (b) Women (1) (2) (3) (4) In school Unempl. Disability Soc. ass. ln(prt) (1.90) (0.91) (-0.05) (1.21) LBW (-0.56) (-0.79) (1.17) (0.82) LBW ln(prt) (-0.41) (1.39) (0.41) (0.40) Observations t statistics in parentheses p < 0.1, p < 0.05, p < 0.01 Note: Table shows selected coefficients from estimation on (1) on persons in the affected municipalities, by gender. Models include municipality fixed effects, year dummies, age (5- year dummies) and county level linear time trends. ln(prt) is the log of the payroll tax, LBW is an indicator of low birth weight (< 2500 grams). Standard errors clustered at the municipality level. 12

13 We have found that the resulting higher labor costs had an additional effect on individuals with low birth weight. Compared to their normal birth weight peers, they experienced larger reductions in earnings and labor force participation. Following the payroll tax hikes, they also had lower rates of enrolling in school, and higher rates of disability insurance receipt, compared to affected individuals with normal birth weight. These findings indicate that higher labor costs could reduce labor demand, in particular for persons with poor ex ante health endowments, who may be less productive employees. That is, we find support for the hypothesis that high labor costs could be keeping workers with poor health out of the labor market. The reversal of the payroll tax harmonization in 2007 makes it difficult to address more long term outcomes. However, the differential impact on program participation reveals a pattern in short term trends that may have long lasting consequences. Going back to school when labor demand drops can be thought of as an investment in human capital, increasing earnings in the long run. Meanwhile, DI is largely an absorbent state, with low transition rates back to labor force participation. The tendency of persons with low birth weight to not go back to school, but rather enroll in DI, may then set them on a path to permanently lower earnings compared to their normal weight peers. This pattern is worrisome too from a fiscal perspective, in particular given the relatively young age of the sample population. References Barth, E., Moene, K. O. & Willumsen, F. (2014), The scandinavian model an interpretation, Journal of Public Economics 117, Bennmarker, H., Mellander, E. & Öckert, B. (2009), Do regional payroll tax reductions boost employment?, Labour Economics 16(5), Bharadwaj, P., Lundborg, P. & Rooth, D.-O. (2013), Health endowments and unemployment during macroeconomic crises. Black, S., Devereux, P. & Salvanes, K. G. (2007), From the cradle to the labor market? the effect of birth weight on adult outcomes, The Quarterly Journal of Economics 122(1), Dale-Olsen, H. (2014), When EU competition rules require non-member state action - labor demand and supply changes in Norway following an imposed harmonization of payroll-tax rates. Duggan, M. G. et al. (2003), The rise in the disability rolls and the decline in unemployment, The Quarterly Journal of Economics 118(1), Kostøl, A. R. & Mogstad, M. (2014), How financial incentives induce disability insurance recipients to return to work, American Economic Review 104(2),

14 Pries, M. & Rogerson, R. (2005), Hiring policies, labor market institutions, and labor market flows, Journal of Political Economy 113(4), Rege, M., Telle, K. & Votruba, M. (2009), The effect of plant downsizing on disability pension utilization, Journal of the European Economic Association 7(4), Reiling, R. B. & Strøm, B. (2015), Upper secondary school completion and the business cycle, The Scandinavian Journal of Economics 117(1), Royer, H. (2009), Separated at girth: US twin estimates of the effects of birth weight, American Economic Journal: Applied Economics 1(1), Saez, E., Matsaganis, M. & Tsakloglou, P. (2012), Earnings determination and taxes: evidence from a cohort-based payroll tax reform in greece, The Quarterly Journal of Economics. Von Wachter, T., Song, J. & Manchester, J. (2011), Trends in employment and earnings of allowed and rejected applicants to the social security disability insurance program, The American Economic Review pp