Employment protection and productivity: evidence from firm-level panel data in Japan

Abstract

Recent developments in the literature on Employment Protection Legislation (EPL) have revealed that changing the stringency of employment protection can lead to extensive consequences outside the labour market, by affecting firms’ production decisions or workers’ commitment levels. This article provides the first empirical evaluation of the comprehensive effect of restrictions on firing employees in Japan, by exploiting the variations in court decisions. We find that judgements lenient to workers significantly reduce firms’ total-factor productivity growth rate. The effect on capital is mixed and inconclusive, although we obtain modest evidence that an increase in firing costs induces a negative scale effect on capital inputs.

Keywords:

• employment protection
• total factor productivity
• capital substitution
• firing cost

JEL Classification::

• J65
• K31
• K41

Acknowledgements

We would like to thank an anonymous referee, Fumio Ohtake, Charles Y. Horioka, Daiji Kawaguchi, Takuji Kawamoto, Masayuki Takahara as well as the seminar participants at RIETI, the spring meeting for the Japanese Economic Association 2008 for their helpful comments. Okudaira acknowledges a grant from the Japan Society for the Promotion of Science.

Notes

¹ For instance, OECD (1999) generated the ordinal index by grading the difficulty in firing procedures or severance payments and found negative but insignificant correlation between the EPL index and the employment rate among young and female workers. Heckman and Pages-Serra (2004) provided a more credible cross-country comparison by constructing the cardinal EPL index for Organization for Economic Co-operation and Development (OECD) and Caribbean countries. Their results indicated that high severance payments moderately decrease total employment. Addison and Teixeira (2001) provided a broad survey of cross-country studies and concluded that strict employment protection has a negative, if not significant, impact on employment rate.

² Besley and Burgess (2004) found that pro-worker amendment of the Indian Industrial Relations Act reduces state output per capita and hinders welfare by increasing the urban poverty. Similarly, Autor et al. (2006) examined the effect of the ‘implied-contract’ exception to the employment-at-will doctrine on employment-to-population ratio and indicated that employment rate falls significantly if state courts accept the exception. Okudaira (2008) found that prefecture employment rate is reduced by approximately 1.5% if a prefecture receives more pro-worker judgements than pro-employer ones in a given year.

³ This reduction is not negligible in size, since, according to their estimates, the industry in a country with OECD-average EPL would experience a 0.08% higher growth rate in total factor productivity if it had EPL as flexible as the US has.

⁴ Autor et al. (2007, pp. F195–196) provides an organized theoretical summary.

⁵ Cabarello et al. (2004) confirms the similar results using sectoral panel dataset across 60 countries.

⁶ See Koeniger (2005), Kanniaine and Vesala (2005) and Samaniego (2006) for the empirical support of this view.

⁷ Koning (2003) makes use of firing information pertaining to soccer coaches in the Dutch Premier League and shows that firing indirectly incentivizes the rest of the workforce.

⁸ The Civil Code provides freedom of dismissal, by stating that ‘‘if the employment is not for a definite period, either party may make a request to terminate the contract at any time…’’ (Civil Code, Art. 627, Par.1). Judges de facto altered this written statute by the Doctrine of Abusive Dismissal in the face of a serious economic downturn immediately after the Second World War, when the cost of being fired was exceptionally high for workers. See Sugeno (2002, pp. 473–93).

⁹ With regard to the second prerequisite mentioned above, for example, the ‘scope’ of a labour contract is one of the questions at issue. The Tokyo District Court has tended to define the relatively narrow scope of a labour contract by requiring firms to reallocate workers only within an affiliated company and not across all related companies (Saitoh v. Chase Manhattan Bank, Rodo Hanrei 63, Tokyo D. Ct. Feb. 27, 1992). Another issue is the legitimacy of hiring new workers immediately before or after the adjustment dismissal. The case of Uenishi v. Meiji Shoin (779 Rodo Hanrei 27, Tokyo D. Ct. Jan. 12, 2000) recognized the legitimacy of the second prerequisite; however, despite this, firms hired new employees around the time of the adjustment dismissals. This was considered to be a radical judgement for relaxing employment protection (Ukai, 2001).

¹⁰ See the Section ‘The Index for Employment Protection’ for details.

¹¹ The specification of our model relies on those from Autor et al. (2007) and Fukao et al. (2006).

¹² The Basic Firm Survey is conducted by the Ministiry of Economy, Trade, and Industry (METI) of the Japanese government. This dataset is often used in published studies (e.g. Kawaguchi, 2007; Morikawa, 2011).

¹³ The original data source is taken from Ohtake (2004) (Okudaira, 2008).

¹⁴ There are obvious reasons why the accumulation commences in 1950. First, the Allied High Command introduced a new set of labour regulations between 1945 and 1947, immediately after the Second World War. The second reason is more obvious: JIS has few adjustment dismissal cases filed prior to 1950.

¹⁵ While the variable construction procedure in Okudaira (2008) is rather similar to the one in Besley and Burgess (2004), Okudaira (2008) has an implicit but important modification. In the Indian case of Besley and Burgess (2004), they accumulated state-specific amendments to the Indian Industrial Relations Act, a written law. The amendment persists within the state forever unless it is abolished. On the other hand, CourtDecision represents the accumulated information of agents regarding judicial environments, including judicial decision standards or judges’ discretion exercised thus far within a prefecture.

¹⁶ Table 1 in this article and Fig. 1 in Okudaira (2008) exhibit this possibility.

¹⁷ Okudaira (2008) shows the OLS results remain the same even in the Instrumental Variable (IV) estimations.

¹⁸ The level of uncertainty in prefecture real gross production is estimated by running a rolling regression of AR(1) with a 10-year sample window for each prefecture 

where indicates real gross prefecture production. The level of uncertainty is defined as the SE of the regression above. This method is followed by Ghosal and Loungani (1996) and Ogawa and Suzuki (2000).

¹⁹ Unfortunately, our data contain no information about firms’ locations prior to our sample period. Thus, we cannot rule out endogeneity in firms’ decisions to enter or shift locations prior to 1994.

²⁰ In order to account for the firm size, we also conducted the similar analysis for the change in investment-sales ratio (I/Y), and obtained the same results. Estimation table is available upon request.

²¹ The instruction in this section is adapted from Fukao et al. (2006).

²² English instruction is available at http://www.rieti.go.jp/en/database/d05.html.