مالیات حقوق و دستمزد، حقوق و دستمزد و اشتغال: شناسایی از طریق تغییر سیاست

This paper investigates the effect of changes in payroll taxes on wages and employment in Argentina. The analysis, based on administrative data, focuses on the impact of a series of major changes in payroll taxes which varied across geographical areas. This setup offers two main advantages over previous studies. First, using longitudinal data, the variation in tax rates across space and time provides a plausible source of identification of their effects on employment and wages. Second, the use of legal tax rates for each area at each point in time provides a remedy for the measurement error bias raised by the use of empirical rates constructed from observed tax and wage bills. Once this bias is accounted for, the results indicate that changes in payroll tax rates are only partially shifted onto wages, and they point to the absence of any significant effect on employment.

The appropriate level of payroll taxes and their influence on labor markets are hotly debated issues (see, among others, Nickell and Layard, 1999). While such taxes usually constitute an important source of government revenue, they drive a wedge between the cost of labor for a firm and the net wage of the worker, and may therefore have distortional effects on the functioning of the labor market. The introduction of a payroll tax implies a downward shift in the labor demand schedule equivalent to the amount of the tax, and standard partial equilibrium incidence analysis states that the extent of shifting from employers to workers depends on the elasticities of labor demand and supply. However, taxes and social security contributions do not necessarily reduce workers' perceived income. Social security contributions may also be regarded as deferred consumption when they take the form, for example, of contributions to public pension programs (Summers, 1989). The introduction of a new payroll tax will presumably translate into an outward shift of the labor supply curve, thereby increasing the negative effect of the tax on wages but reducing its impact on employment (Gruber and Krueger, 1990). Such adjustments could be prevented, however, through the use of bargaining systems that preclude downward adjustments in wages (Dolado et al., 1996). In the case we study in this paper, however, it is unlikely that workers would perceive any significant change in their permanent income through future payroll tax induced revenues. Thus, there would only be a downward shift in the demand schedule as a result of the changes in payroll taxes studied in this paper. The relative levels of labor demand and supply elasticities, the presence of offsets, and the resulting incidence of the payroll tax are, ultimately, empirical questions, although there is a relatively small amount of definitive evidence on the subject. Earlier studies include Brittain's (1971) cross-country analysis, which finds full shifting of the tax to the worker's remuneration, and Holmlund (1983), which finds partial shifting based on time series evidence from the Swedish economy. Most of the shortcomings of cross-country and time series studies are overcome by Gruber's (1997) influential study of Chile's 1981 major social security reform and the resulting reduction of payroll taxes by around 25 percentage points. This study, based on a panel survey of manufacturing plants, compares wages and employment before and after the reform and finds evidence of a full shifting of taxes to wages, with no significant employment effects. Gruber's (1997) data, however, contain no information on the statutory tax rates applied to each company, which are conjectured to have changed across the board for the whole country at one point in time. His study relies on firm-specific empirical tax rates that are calculated by dividing total tax payments by wages. Despite the advantage of having information on the actual tax liability for every firm in the panel, any shock and/or measurement error in wages (the dependent variable) will be reflected as a spurious correlation in the regression results, since the tax rate is by definition a function of wages. Our study exploits a series of major policy shifts in labor taxation which were introduced in Argentina during the period 1995–2001, following a social security reform and other market-oriented changes in the economy. The setting, a middle-income developing country in South America, is reminiscent of Gruber's (1997) Chilean study. However, the Argentine policy reform varied by region, while the Chilean study and later contributions to the literature (Kugler and Kugler, 2003) are based on uniform economy-wide changes or on firm-varying tax levels (Anderson and Meyer, 1997 and Murphy, 2007). Starting from an almost uniform national payroll tax level, the Argentine reform introduced a wide range of rates that varied by geographical area. Moreover, the full set of region-specific legal tax rates were reconstructed from the relevant laws, executive orders (“decretos”) and the software created by the tax authorities which is used by firms to compute tax liabilities.1 Full information is available on the exact legal tax rates applying to firms in each of the regions of the country, which eliminates the problems of spurious regression that might arise from purely observational tax rates.2 These three characteristics (geographical variation, time variation and the availability of exact legal tax rates) constitute the backbone of the identification strategy pursued below. In the spirit of Gruber and Krueger (1990) and of Besley and Burgess (2004), the unit of analysis is a geographical area, and the study is based on an administrative panel dataset containing monthly aggregates of payroll, tax and employment figures. The following section presents a brief account of the institutional setting for the Argentine labor market and its reform during the period under study. Section 3 describes the data and the construction of the variables used in the analysis. Section 4 presents the main empirical results, and Section 5 describes a series of robustness checks. Conclusions follow.

This paper studied the relationship among payroll taxes, wages and employment by focusing on geographical variations in policy changes in Argentina. A distinguishing feature of this study is the availability of exact geographic-specific tax rates. The results indicate that a significant measurement error is carried over from the dependent variable (average wages) to the tax rates constructed by dividing tax payments by wages. The resulting coefficients in this study are roughly twice as high when using constructed rather than legal tax rates. Once this bias is accounted for, the results indicate that changes in payroll tax rates are only partially shifted onto wages, with estimates ranging between 0.4 and 0.9% per percentage-point reduction in the tax rate. They also point to the absence of any significant effect on employment (see Gruber, 1997 and Nickell and Layard, 1999). The period under study includes reductions and increases in payroll tax rates. When these episodes are analyzed separately, the results indicate that only reductions in taxes have a significant impact on wages, although the statistical power of this analysis is limited due to the shorter amount of time available in the data after the tax increase. The presence of full or partial shifting with no employment effects may be explained by a combination of steep (inelastic) labor demand and supply functions, which result in large price effects and negligible quantity changes (Blundell and MaCurdy, 1999). These implied elasticities might result from the limited mobility of the labor force and the regional structure of the labor market in Argentina, since workers do not seem to react to changes in wages.21 Regional labor mobility is limited in Argentina: as noted by Galiani and Nickell (1999), the rate of migration is 1% per year compared to 3% in the U.S. Regional migration is driven mostly by wage differentials, and it can be hampered by transaction costs. These transaction costs are higher in developing countries: there is limited availability of rental housing (because of lower enforcement of property rights), and credit scarcity restricts the mortgage market. Moreover, the Center and South of the country have had significantly higher wages and standards of living than the North, and these wage differentials have become strong pull factors for workers at least since the 1960s. These differences in levels are very large, and dominate any wedge that might be introduced by differential changes in regional tax rates. While the explanation of inelastic supply and demand is consistent with competitive models of wage determination, the presence of shifting only when taxes are reduced might also be indicative of the presence of downward wage rigidity. The main argument justifying the tax cuts introduced in 1995–1999 in Argentina (namely, that lower labor costs would increase employment) was not borne out by events. The wage gains, though significant (with estimates ranging between 0.4 and 0.9 percent per percentage-point reduction in the tax rate), are relatively minor when compared to the historical fluctuations of real wages in Argentina (Galiani and Gerchunoff, 2003). However, the tax cuts did have a significant effect on Government finance: payroll tax collection as a percentage of total wage income (considering the whole country, but excluding public sector and agricultural workers) fell by almost half from the mid 1990s to the early 2000s.