Economics Letters 90 (2006) 328 334 www.elsevier.com/locate/econbase Employment protection: Do firms perceptions match with legislation? Gaëlle Pierre, Stefano Scarpetta T World Bank, 1818 H Street NW, Washington, DC 20433, United States Received 11 January 2005; received in revised form 20 July 2005; accepted 30 August 2005 Available online 27 December 2005 Abstract Drawing from harmonized surveys of firms around the world, we compare employers responses with actual labor legislation. Employers concerns about labor regulations are closely related to the relative stringency of labor laws. Medium and large firms, as well as innovating firms, are those most negatively affected by onerous labor regulations. D 2005 Elsevier B.V. All rights reserved. Keywords: Employment protection legislation; Enterprise surveys JEL classification: J23; K31 1. Introduction Over the past decade, there has been a heated debate on the costs and benefits of employment protection regulations. These regulations have been introduced with the aim of enhancing workers welfare and improving working conditions, but if too onerous they may raise labor adjustment costs and affect labor market outcomes. Theoretical models show that employment regulations constrain both layoffs and hirings, but these models do not provide conclusive answers regarding the aggregate effects T Corresponding author. Tel.: +1 202 458 1119; fax: +1 202 522 7247. E-mail address: Sscarpetta@worldbank.org (S. Scarpetta). 0165-1765/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.econlet.2005.08.026
G. Pierre, S. Scarpetta / Economics Letters 90 (2006) 328 334 329 on employment or unemployment. 1 Empirical analyses for OECD countries have also provided varying results, whether focusing on de jure legislation (Lazear, 1990; and OECD, 1999, 2004 for a survey) or businesses perceptions (e.g. Freeman, 2001; Di Tella and MacCulloch, 2005). The evidence in developing countries is even less conclusive given the lack of enforcement of labor legislation and the sizeable informal economy (Heckman and Pagés, 2004). This paper compares summary indicators of the stringency of employment regulations with the responses of employers on the constraints imposed by these regulations using a sample of about 17,000 firms in 47 developing and emerging countries. These data allow shedding light on two related questions: Do employers in countries with relatively more stringent employment protection regulation tend to report these regulations as more constraining for their operation and growth plans? Which are the firms more likely to report labor regulations as particularly constraining for their operation? 2. Data Using an in-depth review of laws and regulations around the world (the World Bank s Doing Business database) we construct summary indicators of the de jure stringency of employment protection legislation (EPL). 2 Regulations of regular employment include: (i) procedural requirements; (ii) notice period for dismissal; and (iii) severance payments. Indicators of the stringency of the regulation of temporary employment refer to: (i) whether fixed-term contracts are only allowed for fixed-term tasks; and (ii) the maximum cumulated duration of a contract. These variables have been aggregated into an overall indicator of EPL. 3 This indicator passes simple validation tests: for example it correlates well with other indicators produced by the OECD, arguably the most complete measures available. Employers perceptions about employment regulations are drawn from the Investment Climate Surveys (ICS) 4 of the World Bank. In particular, we have used the responses to the following question: bplease tell us if any of the following issues are a problem for the operation and growth of your business. If an issue poses a problem, please judge its severity on a four-point scaleq. The scale is: 0=no, 1=minor, 2=moderate, 3=major, 4=very severe and a list of 18 issues is proposed, including labor regulations. 5 1 See Addison and Teixeira (2003) for a survey. 2 http://rru.worldbank.org/doingbusiness/default.aspx. 3 The aggregation process follows previous studies (e.g. OECD, 1999, 2004). The synthetic indicator of EPL for permanent employment is constructed as the average of the following variables, each composed of different indicators: (i) Procedural inconveniences including: third party notification and approval; priority rules for dismissal and re-employment; definition of fair ground for dismissal; and (ii) Firing costs including mandated notice period for redundancy dismissal after 20 years of continuous service; severance pay for redundancy dismissal. The synthetic indicator of EPL for temporary employment is the simple average of the following variables: (i) Restrictions on the use of fixed-term contracts; and (ii) Maximum duration of fixed-term contracts. The overall index is the simple average of the temporary and permanent employment indexes. All the variables range between 0 and 1. The sample of countries used to construct these variables consists of 83 industrial, developing and emerging economies. See Pierre and Scarpetta (2004) for more details. 4 http://iresearch.worldbank.org/ics/jsp/index.jsp. 5 In a sensitivity analysis we also used the responses to the World Business Environment Survey, which cover a larger number of countries (about 80) but a smaller sample of firms in each country. The results are similar to those presented in this paper.
330 G. Pierre, S. Scarpetta / Economics Letters 90 (2006) 328 334 3. Modeling firms perceptions We assume that the employers perception of the strictness of labor regulations is a continuous latent variable y* which represents utility or preferences that are comparable across individuals. These preferences ( y*), however, are not observable, and we use the response to the Investment Climate Survey ( y), which we take as the manifestation of these preferences. Using a generalized ordered logit model, 6 we regress the perception of labor regulations on our index of strictness of de jure regulations, and a set of control variables. The generalized ordered logit model estimates a set of coefficients (including one for the constant) for each of the m 1 points at which the dependent variable can be dichotomized. From this set of k coefficients (B k ), using the logistic cumulative distribution, it is straightforward to derive formulas for the probabilities that y will take on each of the values 0, 1,..., m: Py¼ ð 0Þ ¼ Fð X b 1 Þ Py¼ ð jþ ¼ F X b ðjþ1þ F X b j j ¼ 1; N ; m 1 Py¼ ð m Þ ¼ 1 Fð X b m Þ where b is a K 1 vector and X contains K explanatory variables. Two sets of basic explanatory variables are considered: (i) firm s characteristics; (age, size of firm, and ownership); (ii) external conditions, for example, enforcement of regulations or economic conditions (captured through the country s income level and region dummies). Firm s perceptions are also likely to depend on firms characteristics. 7 We capture this in two ways: (i) by adding the innovating history of the firm (whether it upgraded or created a product line in the 3 years before the survey) and (ii) by including the lagged changes in permanent employment. The introduction of actual de jure employment regulations, both as a stand-alone variable and interacted with the firm s behavior, raises an important methodological issue. Employment protection indicators only vary across countries and standard errors can be seriously biased downwards (see Moulton, 1990). We used the standard cluster adjustment made to the variance covariance matrix, which assumes that observations are independent across countries, but not necessarily across firms in the same country, and that observations may not be identically distributed. 8 6 The generalized ordered logit model relaxes the proportional odds assumption of a typical ordered logit model and allows the effects of the explanatory variables to vary with the point at which the categories of the dependent variable are dichotomized (Maddala, 1983, p. 46). 7 Firms perception on labor regulations may also be affected by aggregate economic conditions, over and above individual firm s performance. We test this hypothesis by adding a set of dummies characterizing growth patterns: strong growth (top quartile of the distribution of 150 countries by GDP growth rates over the 2 years preceding the survey); medium high growth (2nd quartile); medium low growth (3rd quartile); and low growth (4th quartile). Our results are robust to the inclusion of these dummies and are available from the authors upon request. 8 We also tested for the hypothesis that the appropriate level of clustering is at the industry-country level, and not the country level. Despite the fact that the sample is reduced because of the lack of information on detailed industries in some surveys, the results are largely unaffected and are available upon request. The cluster adjustment is described in Williams (2000).
G. Pierre, S. Scarpetta / Economics Letters 90 (2006) 328 334 331 4. Results The average marginal effects of the independent variables on the probability that the respondent reports labor regulations as no or as a major or severe are presented in Tables 1 and 2. The simple model in the first two columns of Table 1 shows that, other things being equal, young or small firms appear less concerned with labor regulations, while, older, or medium and large firms are more likely to report that these regulations are an to their business. t surprisingly, firms which are at least partly owned by the government show less concern for labor regulations than those completely private. Large employment adjustments in state-owned enterprises often involve generous packages transferring, at least partially, the adjustment costs to taxpayers. Table 1 Firms characteristics and perception of strictness of labor regulations (marginal effects from generalized ordered logit model) Model 1 Model 2 Model 3 Age (less than 5 years old) 5 to 15 years old 0.018 (0.012) 0.005** 0.012 (0.012) 0.004** (0.001) 0.019 (0.014) 0.006* 16 or more 0.005 0.009** 0.018 0.007* 0.011 0.010** (0.020) Publicly owned 0.022 (0.021) Size (fewer than 20 employees) 20 to 100 employees 0.095** More than 100 employees Country income level (low) Lower middle 0.177** 0.008** 0.014** 0.020** (0.005) (0.020) 0.012 0.071** 0.156** 0.007** 0.011** 0.016** 0.029 (0.024) 0.096** 0.178** 0.011** 0.017** 0.023** (0.005) 0.121* (0.057) 0.021 (0.014) 0.124* (0.054) 0.018 (0.012) 0.126* (0.059) 0.023 (0.016) Upper middle or high 0.243** (0.086) 0.046 (0.033) 0.238** (0.084) 0.039 0.246** (0.085) 0.053 (0.038) Innovator 0.076** (0.015) 0.007** Expanded employment 0.047** (0.011) 0.005** (0.001) Observations (number of countries) 17461 17461 14043 (39) 14043 (39) 17140 17140 Predicted probability at the mean 0.441 0.161 0.440 0.169 0.440 0.161 The dependent variable is the perceived degree of constraint. It is an ordered variable taking values 0 = no, 1 = minor, 2 = moderate and 3 = major or very severe. The table presents the average marginal effects and predicted probability of a firm responding bno Q or bmajor or severe Q to the relevant question (see main text). For example in the first column, the predicted probability that the average firm responds bno Q is 44.1%. The probability that a medium firm reports that labor regulations are bno Q is on average 9.5 percentage points lower than for a small firm. Robust standard errors are in parentheses. All regressions include dummies for the geographical location (regions). Where relevant, base categories of dummies are indicated in parentheses in italics. *significant at 5%, **significant at 1%. All estimations control for country clustering. Innovator = upgraded or created new product line. Data source: Investment Climate Survey, World Bank.
332 G. Pierre, S. Scarpetta / Economics Letters 90 (2006) 328 334 Table 2 Perceptions and de jure labor regulations (marginal effects from generalized ordered logit model) Model 1 Model 2 Model 3 Model 4 Size (fewer than 20 employees) 20 to 100 employees 0.093** (0.021) More than 100 employees 0.178** Overall EPL 0.268* (0.125) 0.013** 0.020** 0.206** (0.064) 0.007 (0.035) 0.119** (0.042) 0.003 (0.005) 0.014* (0.008) 0.094** (0.020) 0.180** 0.014** 0.022** 0.057** (0.016) 0.141** (0.021) 0.008** 0.013** small firms 0.162 (0.145) 0.156 + (0.085) medium firms 0.395** (0.135) 0.250** (0.066) large firms 0.307* (0.124) 0.203** (0.060) Expanding employment 0.032 0.003 + expanding employment 0.268* (0.118) 0.203** (0.060) contracting employment 0.300* (0.137) 0.212** (0.071) Innovator 0.016 (0.034) 0.0003 innovator 0.385** (0.134) 0.253** (0.067) not innovator 0.209 + (0.123) 0.153 + (0.088) Observations (number of countries) 16847 16847 13827 (38) 13827 (38) Predicted probability at the mean 0.437 0.163 0.437 0.163 0.435 0.163 0.437 0.171 Robust standard errors are in parentheses. + Significant at 10%; *significant at 5%; **significant at 1%. All estimations control for country clustering, and control for firm s age, government ownership, region and country s income. Base cases are in italics. Innovator: upgraded or created product line. See also notes to Table 1. Data source: Investment Climate Surveys, World Bank. In the second set of specifications (columns 3 to 6) in Table 1, we add firm performance indicators (i.e., innovating history and employment expansion). As could be expected, innovating firms are more likely to report that labor regulations are a significant source of constraint. Oftentimes, innovation involves the adoption of new technologies that require new skills, and labor regulations, by raising the costs of adjusting the workforce, may make it more difficult or costly to do so. t surprisingly, firms that increased the number of full-time staff are found to be less concerned by labor regulations than others. 9 9 We use the change in employment to characterize the general performance of the firm and, to reduce the obvious endogeneity problem, we use the lagged values of the employment change. Results in Tables 1 and 2 show that all other coefficients are robust to the inclusion of this variable.
G. Pierre, S. Scarpetta / Economics Letters 90 (2006) 328 334 333 When including formal regulations in the specification (Table 2), we find that even controlling for firm characteristics, income per capita and geographical location of the countries, de jure regulations are consistent with employers perceptions: firms facing stricter EPL are more likely to report that regulations are a major to their operation. To assess the influence that firm s characteristics and behavior have on the link between actual regulations and perceptions of regulations, we allow the coefficients of EPL to vary depending on firm size (columns 3 and 4). While all firms perceive regulations to be more constraining in countries with strict EPL, medium and large firms are more severely affected. Small firms are often exempted from certain aspects of labor regulations (Boeri and Jimeno, 2003) or, when enforcement is weak, do not comply, remaining invisible to regulators and inspectors. Larger firms tend to be more sensitive to the strictness of regulation; the estimated average marginal effects would suggest that a 0.2 increase in EPL indicator (about one standard deviation) would increase the probability that medium-sized and large firms report that labor regulations are a major constraint by 36% and 26%, respectively. There is no evidence that firms employment expansion influences the link between perception and EPL (Table 2, columns 5 and 6). By contrast, innovating firms are more likely to report more severe constraints to labor regulations in countries with more stringent EPL, as shown by the significant effects of the interaction terms between innovation and EPL (columns 7 and 8). 10 This may have implications for policy makers: by curbing incentives for firms to innovate and adopt new technologies, onerous EPL may over time negatively affect the aggregate productivity and growth potential of the country. Acknowledgments For helpful comments we are grateful to an anonymous referee and to Mary Hallward-Driemeier, Alan Gelb, Christine Jolls, Carmen Pagés, Vijaya Ramachandran and Warrick Smith, and participants at two World Bank seminars and at the 2nd Joint World Conference of SOLE-EALE, San Francisco, June 3 5, 2005. The views expressed in this paper are our own and should not be held to represent those of the World Bank. References Addison, J., Teixeira, P., 2003. The economics of employment protection. Journal of Labor Research 24 (1), 85 129. Boeri, T., Jimeno, J., 2003. The Effects of Employment Protection: Learning from Variable Enforcement. CEPR Discussion Papers,. 3926. Di Tella, R., MacCulloch, R., 2005. The consequences of labor market flexibility: panel evidence based on survey data. European Economic Review 49 (5), 1225 1259. Freeman, R.B., 2001. Institutional differences and Economic Performance among OECD countries. Paper presented at the Bank of Portugal Conference blabor Market Institutions and Economic OutcomesQ, Cascais, Portugal, June. Heckman, J., Pagés, C. (Eds.), 2004. Law and Employment: Lessons from the Latin America and the Caribbean. National Bureau of Economic Research, Cambridge, MA and University of Chicago and Chicago, IL. Lazear, E.P., 1990. Job security provisions and employment. Quarterly Journal of Economics 105 (3), 699 726. 10 While the coefficients of b expanding employmentq is not significantly different from that of boverall EPL and contracting employmentq, the coefficients of b innovatorq and b not innovatorq are significantly different from each other.
334 G. Pierre, S. Scarpetta / Economics Letters 90 (2006) 328 334 Maddala, G., 1983. Limited-dependent and Qualitative Variables in Econometrics. Econometric Society Monographs. Cambridge University Press. Moulton, B., 1990. An illustration of a pitfall in estimating the effects of aggregate variables on micro units. Review of Economics and Statistics 72, 334 338. OECD, 1999, 2004. Employment Outlook, Paris. Pierre G., Scarpetta, S., 2004. Employers Regulations Through the Eyes of Employers: Do they Matter and How Do Firms Respond to Them? IZA Discussion Paper 1424, December. Williams, R., 2000. A note on robust variance estimation for cluster-correlated data. Biometrics 56, 645 646.