Who Can Reform the Labor Market? IMF Conditionality, Partisanship, and Labor Unions

¹ For a summary of IMF’s advice on labor market issues, see: https://www.imf.org/external/np/exr/facts/labor.htm.

² In fact, implementation rates are lowest for the reforms pertaining to the labor market.

³ I use special interests, organized interests, and organized interest groups terms interchangeably throughout the text.

⁴ Figures A1 and A2 in the Appendix provide an overview of the evolution of labor market conditionality in IMF programs.

⁵ Although the scholarly work on the IMF has documented and studied heterogeneity in the conditionality targets, policy areas, depth, and stringency (Beazer and Woo Citation2016; Caraway et al. Citation2012; Dreher Citation2004, Citation2009; Dreher and Vaubel Citation2004; Dreher et al. Citation2015; Rickard and Caraway Citation2014), the potential consequences of policy heterogeneity for the implementation of reforms have not been systematically studied (Caraway et al. Citation2012).

⁶ I do not claim that IMF programs always lead to positive economic outcomes. Questioning the effectiveness of the IMF policies on labor market is beyond the scope of this study.

⁷ I standardized the Polity2 scores by adding 10 each, and thus it is bound between 0 and 20.

⁸ These partisan differences have been supported by many studies that documented the variations in macroeconomic priorities (Garrett Citation1998; Hibbs Citation1987) and the interaction between partisanship and various domestic institutions such as labor organizational structure and labor market organizations (Boix Citation2000).

⁹ Although Rueda (Citation2007) argues that partisan preferences matter, he shows that Social Democrats are increasingly interested in defending the interests of insiders rather than outsiders.

¹⁰ For instance, the public sector union members in Ireland took to the streets to protest against the public service reform measures and wage cuts planned under the IMF program.

¹¹ This study does not assume that organized interests do not matter in autocracies. Although these groups might also have access to decision makers in autocratic settings through certain clientelist linkages, their effect on average should be larger in democratic settings.

¹² Organized labor has been effective by delaying or preventing labor market reforms in Latin America (Murillo Citation2001; Murillo and Schrank Citation2005) and Europe (Alexiadou Citation2013; Featherstone Citation2008; Featherstone and Tinios Citation2006).

¹³ For instance, the German social democratic party initiated a major labor market reform in 2003, cutting unemployment benefits and changing employment protection rules (Lunz Citation2013).

¹⁴ Similarly, Garrett (Citation1998) shows that policy reform is most likely when left-wing governments work together with strong organized labor or where right governments are faced with weak/decentralized labor.

¹⁵ This perspective is especially true for the IMF. Recent examples from Europe, such as in Italy and France, highlight the left’s willingness to reform the labor market (Ekathimerini Citation2016; Henley and Inman Citation2016; Reuters Citation2015).

¹⁶ This opposition may also arise despite the knowledge of or due to uncertainty regarding the expected long-term benefits of reform (Fernandez and Rodrik Citation1991).

¹⁷ The logic is explained by Cho (Citation2014) by referring to the “Nixon goes to China” idea.

¹⁸ The theory presented here might not be consistent with some institutional settings such as corporatist structures in which labor market reform can be achieved through other means such as social partnership/social dialogue between organized labor, business interests, and government. These processes are especially common in advanced industrial societies such as Scandinavian countries. However, the countries included in the IMF sample do not have such features.

¹⁹ Median voter theory also suggests that, when labor is militant, it is easier for the left government to move to center and form a grand coalition against the (more radical) labor.

²⁰ The literature on the IMF, which focuses on the effect of electoral concerns and timing of elections, shows ambiguous results (Arpac et al. Citation2008; Dreher Citation2003).

²¹ This program had achieved a significant fiscal consolidation corresponding to 5% of the GDP in its first year. The Fund praised the government’s willingness and effort and defined this as an impressive achievement by international standards (IMF Citation2010b; Wyplosz and Sgherri Citation2016).

²² Similarly, the coalition partner, PASOK, stalled implementation in various areas that affected their core constituency, the public sector workers.

²³ There are 11 different policy areas: General Government (including taxation, expenditure reforms and customs); Central Bank; Public Sector (wages and employment); Social Policy (pensions, health and education expenditure/systems); Public Sector Privatization and Pricing; Financial Sector Regulations and Privatization; Exchange Rate Regimes; International Trade; Labor Market; Statistics; Other reforms (governance, corruption, etc.).

²⁴ One can question the inclusion of privatization reforms. However, these reforms generally require changes in the employment conditions in the publicly owned enterprises and lead to lay-offs. Thus, they have a significant effect on workers in the public sector.

²⁵ Examples of labor market conditionality in the IMF programs include: “Adoption of a decree that eliminates the portability of the seniority bonus and submission to parliament of an amendment to the labor code that facilitates work outside regular hours” (Bulgaria 2004 SBA program); “Limit the wage increase for budgetary workers in November 2008 to 5.3 percent” (Belarus 2008 SBA); “Adopt a comprehensive pension reform that reduces the projected increase in public spending on pensions over the period 2010–16 to 2½ percent of GDP” (Greece 2010 SBA program).

²⁶ The focus in this paper is on a typical IMF borrower, which has an average polity score of (at least 7). This reflects the changes within the last two decades with increasing democratization.

²⁷ The IMF non-concessional loans are provided through Stand-By Arrangements (SBA), the Flexible Credit Line (FCL), the Precautionary and Liquidity Line (PLL), and the Extended Fund Facility.

²⁸ The IMF concessional loans are provided through Poverty Reduction and Growth Trust (PRGT), Poverty Reduction and Growth Facility (PRGF), the Standby Credit Facility (SDF) and the Rapid Credit Facility (RCF).

²⁹ Concessional programs might include labor market conditions favorable to labor (or some parts of the labor force). For instance, the Armenia program in 2009 included a condition on strengthening the social safety nets to reduce the social impacts of the program. Similarly, Ghana’s 2009 PRGF program, although it required public sector hiring freeze, also provided exemptions to health and education employees. Moreover, some IMF public sector and labor market conditions in concessional programs aim to develop capacities such as increasing the efficiency of public administration (Honduras 2009 PRGF) and initiating poverty alleviation program targeted on households living in extreme poverty (Grenada 2006 PRGF).

³⁰ While 76% of all non-concessional program/years has at least one labor market condition, 71% of all concessional program/year observations have at least one labor market condition. The average is higher for non-concessional (3.31) than concessional programs (2.26). Lastly, non-concessional programs rely more heavily on fiscal measures and labor market conditions than social policy measures, which are more common in concessional programs.

³¹ This is often the case in IMF programs since IMF might decide to cancel/exclude some conditions from the program or include new conditions. This dynamic approach is better suited to analyze the implementation process than the design stage, which can be seen as a negotiation prior to the conclusion of the initial letter of intent. However, implementation takes place through a couple of reviews and might require constant negotiation of the terms.

³² I also have two alternative codings of these half-met conditions. 1) coded as met 2) exclude them from the analysis. Because there are only a few half-met conditions, the findings do not change with these alternative specifications.

³³ One critique to strikes as a measure of strength of labor is that it might actually be a sign of a weakening labor movement. However, even if it is considered as a sign of weakening labor, it still suggests that the labor can organize and initiate strikes at the national level, compared to the unions that cannot strike. This creates a politicized environment in which reforms become more visible, costly and contentious. In summary, strikes act as a measure of cost of reform, increasing the visibility of the reforms and putting extra pressure on the policy makers even if the labor might strike as an act of desperation.

³⁴ To overcome the concerns that exclusion of union density might result in omitted variable bias, I also present models with union density in the Appendix, Table A6.

³⁵ The executive here refers to a president in presidential regimes and the prime minister in parliamentary regimes.

³⁶ I rely on different sources for the ideology variable because the World Bank database has many missing variables and does not enable a comprehensive analysis.

³⁷ The sample includes both presidential and parliamentary regimes. Thus, it is possible that the sample in this study also includes coalition governments. This study does not necessarily theorize on the type of the government, or the number of parties in the coalition. However, to control for any possible effect of coalition partners, I also include government ideological fractionalization as additional robustness controls. Moreover, the political constraints variable also measures veto players within the executive branch of the government.

³⁸ Because the data set actually codes the date of implementation as well, I am able to distinguish the conditions that are implemented before and after elections.

³⁹ In alternative models, I also use veto players and polarization variables from the World Bank’s Database on Political Institutions. These results are available upon request.

⁴⁰ I provide a complete list of missing observations for each variable and analysis in the Appendix, Table A2.

⁴¹ I created 20 imputed data sets with 100 iterations using multivariate normal distribution (MVN).

⁴² I also use the percentage of programs rather than the total number of programs both in the global and regional level. The results, which are available upon request, are robust to these specifications as well.

⁴³ Although not discussed in detail here, these models also provide substantively different insights. While the fixed effects model suggests that these findings hold within each unit, which is for countries, random effects model suggests that these effects are on average more likely in a global sample of countries with similar characteristics. However, a Hausman test does not confirm that we should use fixed effects model.

⁴⁴ Surplus labor is calculated as the (working age population minus students enrolled in secondary education minus students enrolled in “post-secondary” education) minus (laborforce/the working age population). Because PLP measure is only available until 2000, I use surplus labor. The correlation between PLP and labor surplus is 62 percent.

⁴⁵ These are based on the authors’ own coding of the IMF conditionality from the MONA database.