The effects of hiring and firing regulation on unemployment and employment: evidence based on survey data

Abstract

We use the results of surveys among senior business executives to measure the strictness of hiring and firing regulations. The survey data are more likely than objective indicators (used in almost all previous studies) to correctly capture the de facto strictness of these regulations and their relevance to the performance of the labour market. Using data from 19 industrial countries for the period 1992 to 2002, we find that more flexible regulations are likely to lower unemployment and to increase employment rates. While the effects on the general population appear to be modest, the effects on female, young and low-skilled workers seem to be substantial.

Notes

¹See, e.g. Lazear (1990), Addison and Grosso (1996), Scarpetta (1996), Elmeskov et al. (1998), Feldmann (2003, 2005), International Monetary Fund (2003), Heckman and Pagés (2004), OECD (2004a), Baker et al. (2005), Nickell et al. (2005) and Bassanini and Duval (2006).

²Over time, there have been some changes to the World Economic Forum's surveys that are of minor importance to our analysis. First, between 1989 and 1995 the surveys were conducted in collaboration with the Institute for Management Development, Lausanne, and between 1996 and 2001 in collaboration with the Center for International Development at Harvard University. Second, the number of countries covered has increased steadily from 36 in 1992 to 80 in 2002. (Due to a lack of data on some of our control variables, our empirical study covers only 19 industrial countries. See text and Appendix A for details.)

³As higher marks on the EOS scale indicate more flexible regulation, we label our variable of interest ‘flexible hiring and firing regulation’.

⁴The Hausman test indicates that the random effects estimates may be biased (Table 1).

⁵A second version of the EPL indicator additionally includes specific requirements for collective dismissals. However, data for this broader indicator are only available since the late 1990s. This does not pose a major problem, though, as specific requirements for collective dismissals do not play a major role. Indeed, as the OECD (2004a, p. 72) has demonstrated, taking account of these specific requirements in the overall measure of EPL strictness does not affect cross-country comparisons much.

⁶This variable is meant to test Calmfors and Driffill's (1988) hump hypothesis, according to which unemployment (employment) will be comparatively high (low) if wages are negotiated at the industry level.

⁷According to previous studies on both transition and industrial countries, female employment rises with rising income per capita (Feldmann 2005, 2006). These studies also indicate that countries with a higher percentage of young people enrolled in tertiary education have a lower percentage of young people in employment.

⁸While controlling for the effects of most other major labour market institutions, the business cycle, the level of economic development and unobserved country effects goes a long way to avoid omitted variables bias, endogeneity still may be a problem for estimation because changes in unemployment and employment rates may lead to changes in the (perceived) strictness of hiring and firing regulation. Unfortunately, due to a lack of instruments we are unable to directly address the reverse causality problem. (This is in line with the previous literature, which does not use instruments either.) However, it is unlikely that reverse causality is relevant in our case because in the regressions presented in Table 1, the coefficient on output gap is either statistically insignificant or has the “wrong” sign. [Higher output gaps are strongly correlated with lower unemployment and higher employment rates (Tables 2 and 3).]

⁹To save space, the estimates for the controls from the random effects FGLS and the pooled OLS regressions are not presented in Tables 2 and 3.

¹⁰We also checked the robustness of our results by excluding from the sample statistical outliers, or any particular country, or any random draw of 10% of observations. None of these checks had any noticeable impact on the coefficient on our variable of interest (results not reported here).

¹¹Important studies on the effects of labour taxes include, e.g. Daveri and Tabellini (2000) and Prescott (2004).

¹²See, e.g. Jenkins and Garcia-Serrano (2004), Lalive and Zweimüller (2004) and Nickell et al. (2005).

¹³As mentioned in Section I, the evidence from previous empirical studies, which almost exclusively use objective indicators, so far is mixed. By contrast, our results are less ambiguous. The main reason for this difference may be that our survey-based indicator is more likely than objective indicators to correctly capture the de facto strictness of hiring and firing regulations and their relevance to the performance of the labour market (Section II).