Promoting jobs and protecting people?

Promoting jobs and protecting people?

Youri ten Hoeve | s2910802

y.ten.hoeve.1@student.rug.nl

Supervisor: prof. dr. van Ark

MSc International Economics & Business

University of Groningen | Faculty of Economics & Business

Abstract

This thesis aims to uncover whether the efficacy of wage setting institutions evolves with economic development. It is found that on net stronger wage setting institutions result in a lower labour market participation rate and/or a more egalitarian income distribution in developing countries. However, these conclusions still leave room for ambiguity as wage setting institutions are subject to between-country heterogeneity. This implies that each country demands its own optimal institutional arrangement. In particular, the size of the informal sector is found to weaken the functioning of wage setting institutions, since this imperfection reduces the number of workers represented by such institutions, diminishing their bargaining power.

Keywords

2

1. INTRODUCTION

The labour market, by definition, is no ordinary market. Although the conventional economic dynamics of supply and demand are at play, the resulting equilibrium is not one of utility maximisation. The stark difference between the labour market and product markets is the nature of consumers. Essentially, in labour markets, firms are the consumers and they are not driven by a conventional utility curve; rather they are driven by profit. The demand dynamics in a competitive labour market are derived from the marginal product of labour. This implies that a profit-maximising firm will hire workers until the marginal product of labour, y, equals the marginal cost of labour, the wage (Boeri & van Ours, 2008). While this economic model starts with the assumption of a perfectly competitive labour market, intuitively, the labour market is not perfectly competitive at all. Workers and firms differ considerably in their degree of information regarding the market. Further, there are frictions, i.e. costs, that hinder the supply and demand dynamics to flow naturally towards the market equilibrium.

This is where wage setting institutions step in. Wage setting institutions are institutions that have a direct effect on the price of labour. As the name suggests, they are either setting a wage directly, such as the minimum wage, or involved in the process of wage determination, such as collective bargaining agreements and labour unions. As early as the emergence of the Industrial Revolution, scholars have analysed industrial relations. Even Adam Smith himself, in his seminal work An Inquiry into the Nature and Causes of the Wealth of Nations, argued that a worker has weak bargaining power relative to its employer and that this will likely cause issues in terms of wage determination (Smith, 1776). It is exactly this concept of bargaining power that has continuously been renowned as the catalyst for wage setting institutions to emerge. The mission statement of the International Labour Organisation, henceforth ILO, captures this sentiment perfectly: “Promoting jobs, protecting people” (ILO, 2019). Ultimately, this statement conveys two broad reasons for wage setting institutions to exist in the first place, the first is efficiency, which implies that such institutions are created to improve the functioning of the market, thereby promoting jobs. The second is equity, which implies that such institutions should make the relations in a labour market more egalitarian, thereby protecting people. As a consequence, wage setting institutions are theoretically expected to create welfare.

3 At the same time, the benefits of wage setting institutions regarding equity become more significant with development. In labour markets that are characterised by market imperfections, wages are not purely determined by supply and demand forces (Booth, 2014). In such markets, the degree of information asymmetry may disproportionally allocate bargaining power between supply and demand, resulting in an inegalitarian wage distribution. Thus, the efficacy of wage setting institutions regarding equity is expected to improve with economic development, due to them covering a larger share of workers. Similarly, in a more developed environment, such institutions are expected to be better managed, which implies that they are expected to be more effective.

Initially, the effects of regulation and organisation of labour sets to increase the competitiveness of developing labour markets by combating market imperfections and attracting people to the formal labour force. If a labour market develops, the impact of wage setting institutions on efficiency is expected to diminish as the market will theoretically become more perfect. At the same time, equity is expected to become more relevant, as wage setting institutions are representing a larger portion of workers and therefore have a larger influence on the process of wage determination. Thus, the efficacy of wage setting institutions is expected to follow a U-shape distribution with regards to development.

“Promoting jobs, protecting people”. As mentioned before, this statement perfectly conveys

why wage setting institutions came to be. It is exactly this message, that I aim to research empirically. More specifically, the research question that is central to this thesis is:

Do wage setting institutions promote job creation (i.e. increase efficiency) and protect people (i.e. foster equity) in developing countries?

Zooming in on developing countries is relevant as it provides context for the more complex U-shape distribution discussed above. This thesis aims to analyse a part of this U-U-shape distribution. Its purpose is to examine the importance of wage setting institutions in developing labour markets given the fundamental role of institutions in economic development. Furthermore, there is a myriad of literature and data on labour market institutions in developed countries, whereas for developing countries, research has thus far merely scratched the surface. First and foremost, more attention must be paid to market imperfections, given the large role they play in the labour markets of developing countries (Fields, 2011). Second, the main policy advice since the 1980s has been grounded in the Washington consensus, which focuses on liberalising labour markets and therefore has been critical towards wage setting institutions (Belser & Rani, 2015). In a labour market that is imperfect by definition, I aim to uncover that stronger wage setting institutions have welfare-enhancing effects in terms of efficiency and equity.

4 Chapter 6 presents a robustness check and lastly, chapter 7 will reiterate the implications of the outcomes and conclude.

2. LITERATURE REVIEW

In analysing labour markets, two schools of economics, neo-classical economics and institutional economics, dominate the discussion. In this literature review, these two perspectives are examined first in order to set the stage for a discussion on the role of wage setting institutions in modern labour markets. The focus then shifts to the labour markets of developing countries, which thus far have been underrepresented in academic debate.

2.1 The neo-classical labour market

In the neo-classical view of the world, the labour market is analysed as any other economic model, despite its peculiarities. The neo-classical labour market is built upon the intuition that both workers and firms, i.e. supply and demand, are ‘utility’-maximising (Boyer & Smith, 2001). Essentially, this means that it is assumed that labour markets are perfectly competitive. On the demand side, profit-maximising firms will hire workers up to the point where marginal productivity equals the marginal cost of labour. Theoretically, these supply and demand forces will steer the market towards an equilibrium, in which all workers with a reservation wage larger than the equilibrium wage are employed (Boeri & van Ours, 2008). Introducing wage setting institutions to this model creates a wedge between the reservation wage of workers and the value of a job, which reflects the firms’ willingness to pay for a marginal unit of labour. In essence, this implies that wage setting institutions are market-distorting as their mere existence alters both the prices and quantities for labour. As a result, stronger wage setting institutions raise the prices for labour and reduce the labour supply. In Figure 1, this market-distorting wedge is displayed.

Following this train of thought, it can be reasoned that wage setting institutions theoretically reduce the size of the labour market and therefore make it less efficient. Less workers will be employed as it has become marginally more costly for firms to hire an additional worker, which is reflected by a left-ward shift of labour supply in the model. Not only do wage setting institutions make the market less efficient, they also have a welfare-reducing effect, since the aforementioned wedge results in increased unemployment which is detrimental to personal welfare (Deaton, 2012).

5 If wage setting institutions are market-distorting by definition, why do they exist? Given that Hicks (1932) argued that in a competitive labour market, supply and demand forces will steer the market to an efficiency optimum and thus interventions, be that by governments or organised labour movements, are undesirable. Moreover, neo-classicists would classify unions and similar groupings of labour supply as cartels, as they possess considerable market power over buyers in the labour market (Kaufman, 2000). Theoretically, this is in line with the wedge that wage setting institutions create, in the sense that such movements exert their power to gain a higher price for their members. Ultimately, if this higher price exceeds the equilibrium wage by a considerable margin, the market is condemned to be inefficient.

2.2 Institutionalists’ view of wage setting institutions

In contrast to the neo-classicists, institutionalists put an emphasis on the attainment of societal utility. This difference in interpretation of utility-maximising behaviour suggests a different role for wage setting institutions. In the view of institutionalists, the purpose of such institutions is to protect the social value of labour (Kaufman, 2000). Early voices in the institutionalists’ school of thought argued that the economic purpose of wage setting institutions is to redistribute wealth and improve wages and working conditions (Commons, 1913). In a similar vein, workers are not just seen as economic actors with the rationale of maximising their income, they are subject to an imperfect market with superior demand forces, due to the inequality in bargaining power. Wage setting institutions are thus on net a positive force as they make the wage determination process more competitive (Kaufman & Lewin, 1998). Alternatively, wage setting institutions act as safeguards against monopsony; which entails a monopoly by employers1.

The core idea of the neo-classical labour market is that wage setting institutions inhibit the market from reaching optimal outcomes. The simplicity of this idea marks its ease of application in policy, as its message of deregulation is easy to get across especially given its presumed positive effects on job growth (Bradley & Stephens, 2012). The political context in which such policies come to be is relevant though, as the ‘varieties of capitalism’ literature posits that stronger wage setting institutions may overall be more beneficial in economies that are characterised by a higher degree of coordination (Hall & Gingerich, 2009). Zooming out, Nickell (1997) points out that labour market institutions do not uniformly inhibit employment creation. Some labour market institutions were actually found to not have any influence on unemployment, despite the claim that they were creating rigidities suppressing perfect competition. The hypotheses advocated by neo-classicists institutionalists are not mutually exclusive, as Bradley & Stephens (2012) found that hypotheses derived from both schools of thought hold merit in OECD countries.

Even though the cross-country variation suggests that the influence of labour market institutions is not uniform, the root of the debate resides in the sources of utility, be that individual or societal. One argument that seems to be consistent across the neoclassical and institutionalist views is that both aim to abolish market imperfections. However, the neo-classicists view wage setting institutions as the main market imperfection, whereas the institutionalists argue that the

6 market is inherently imperfect (Kaufman, 2012). This debate perfectly sets the stage for the two goals of labour market institutions, efficiency and equity (Boeri & van Ours, 2008).

2.3 Wage setting institutions and efficiency: promoting jobs?

During the late twentieth century, economists pointed towards the heavily regulated European labour markets, and the relative strength of labour unions, as the source of the alarming unemployment rates. They argued that the United States, with its far more flexible labour market, had a far more efficient labour market (OECD, 1994). There are multiple issues with this claim: First, it regards the labour markets of European countries as uniform and second, it does not allow any heterogeneity within the concept of labour market institutions. Rovelli & Bruno (2008) address this first issue, by pointing out that there is no single model that represents the labour markets of individual European countries2. Rovelli & Bruno (2008) find that the EU27 differ in terms of policy generosity and therefore have different employment outcomes. More specifically, they found that higher rates of employment are associated with higher expenditures on labour market policies.

If one were to take a step back and conceptualise the labour market as inherently imperfect, wage setting institutions are seen as levelling the playing field, rather than restricting market forces to create the ‘perfect outcome’ (Kaufman, 2012). However, it can also be argued that wage setting institutions bolster efficient bargaining. According to Freeman (1993), the Coase theorem3 can be applied to wage setting institutions. This argument is in line with the aforementioned proposition that labour markets differ across countries and therefore, the effects of wage setting institutions differ significantly across the board. Striving for a unilateral model is in stark contrast with the idea pushed by Coase, who argued that dependent on context, a specific set of institutions will result in minimisation of transaction costs (Coase, 1992). Given that efficient bargaining may minimise transaction costs in the labour market, wage setting institutions are thus expected to allocate resources optimally. Therefore, Freeman (2005) argues that the claim that wage setting institutions inhibits efficiency is far from conclusive, since the data presents mixed results with regards to its effect on aggregate efficiency.

Additionally, there is a stream of literature that argues that if imposed in an imperfect market, wage setting institutions, such as a minimum wage, can improve the efficiency of the labour market (Agell, 2003). This intuition is in line with the idea of monopsony, which implies that firms have considerably more power relative to workers in the labour market. Furthermore, it is suggested that a minimum wage results in a more educated workforce (Agell & Lommerud, 1997; Ravn & Sørensen, 1997). In this sense, it raised labour productivity and efficiency in the long term, despite creating unemployment for the uneducated in the short term. In a similar vein, Kahn (2000) finds that labour market institutions result in a greater propensity for both males and females to attend education. A similar result is found for vocational training (Acemoglu & Pischke, 1999). However, the effect of wage setting institutions is unevenly distributed between groups. Neumark & Wascher (2004) suggest, for instance, that a minimum

2 _{The European countries can be organised into four broad groups of social policy models: The Nordics, the Anglo-Saxon} countries, continental countries and the Mediterranean countries (Boeri, 2002).

7 wage is negatively associated with youth employment rates. However, these findings were especially robust in countries with flexible labour markets; implying that employment protection mitigates this relationship. Other empirical evidence shows that wage setting institutions – in this case minimum wages – have negative effects on employment. More specifically, effects on employment of youth, females and lower skilled were found to be negative. Similarly, workers in small firms seem to be disproportionally affected (Neumark & Nizalova, 2007; Rama, 2001).

In conclusion, on balance the literature posits that wage setting institutions are negatively related to labour market efficiency. Wage setting institutions, for instance, are found to result in higher unemployment (Nickell & Layard, 1999). Furthermore, Bassanini & Duval (2006) found that a reduction in unemployment benefits is associated with a significant drop in the employment rate. So there is merit to the claim that wage setting institutions are distorting the competitiveness and in turn making the labour market more rigid (Baker & Schmitt, 1999; Nickell & Layard, 1999; Neumark & Wascher, 2004). Even though there is plenty of empirical evidence that wage setting institutions are indeed detrimental to labour market efficiency, there are two arguments that put this claim in perspective. First, labour markets are subjects to widely diverging contexts (Rovelli & Bruno, 2008), which implies that there is a myriad of externalities that have some influence on the efficacy of wage setting institutions. Second, the effect of wage setting institutions is not uniform across countries. Ultimately, wage setting institutions on net seem to have a distorting effect on efficiency, but the extent of this distortion does not imply that deregulation is the cure-all solution (Bradley & Stephens, 2012). Therefore, no war of models needs to be fought (Freeman, 1998).

2.4 Wage setting institutions and equity: protecting people?

Similar to its relation with efficiency, the effect of wage setting institutions on equity is ambiguous as well. In the neo-classical model the labour market is assumed to be perfectly competitive, with a homogenous labour force. In reality, workers are not homogenous, due to differences in skill for example, resulting in some workers earning wage premia4. In addition, the information relevant for the wage bargaining process is unfairly distributed to workers vis-à-vis firms. This is exactly why the institutionalists view that the role of labour unions is twofold. As mentioned before, they view that the economic purpose of unions is wealth distribution and the advancement of better working conditions (Commons, 1913). The second role of unions, dubbed its political role5, is that they are a vehicle for democratisation in industrial relations by balancing the bargaining power between firms and workers. (Kaufman, 2000).

Given the multi-faceted purpose established by the institutionalists, it must empirically hold that in countries with stronger wage setting institutions, earnings are more evenly distributed. Literature has established that stronger wage setting institutions are indeed associated with lower inequality in wages, and its effect is as strong as trade and technology measures (Koeniger, et al., 2007). Likewise, Calderón & Chong (2009) find a negative relationship

8 between labour regulations and the income inequality. They found that especially labour union membership, days of maternity leave and the ratification of ILO convention 87 – which allows workers to establish collective bargaining units – are positively influencing an egalitarian income distribution. Interestingly, minimum wage regulations, public employment and social security seem to have a comparatively weaker effect on the income distribution (Caldéron & Chong, 2009).

Freeman (2007) found that countries that rely more on institutions in terms of wage determination are subject to a more egalitarian income distribution. This was found to hold empirically, as, for example, the United States had a considerably higher Gini coefficient than Norway. The relationship between wage setting institutions and equity thus holds at the macro-level. At the micro-level, the data also clearly points towards the equity-enhancing nature of wage setting institutions, since the wage dispersion among union members is lower compared to non-union members (Card, et al., 2004; Freeman, 1984). Nevertheless, this outcome hints at the existence of an insider-outsider problem. If an insider group is relatively small, one can expect that the income distribution of these insiders may become more egalitarian, whereas the outsiders are punished twice: they lose out relative to the insiders, but also do not gain much from the bargaining conducted by labour unions. Aside from a direct effect of wage setting institutions on inequality, there may also be significant indirect effects. Checchi & Garcia-Peñalosa (2008) for instance, posit that stronger wage setting institutions spur investments in human capital; thus reducing inequality in the long run. The effect of wage setting institutions on inequality becomes ambiguous if one considers that a heightened minimum wage will both reduce inequality and increase unemployment (Checchi & Garcia-Peñalosa, 2010). Intuitively, this ambiguity suggests that the concepts of efficiency and equity are interlinked.

Piketty also touches upon the effect of institutions on the distribution of incomes. More specifically, he points out that trends regarding income inequality seem to follow the evolution of the minimum wage (Piketty, 2014). However, he finds that such policies merely influence the bottom of the income distribution, as there is no effect of the minimum wage on top income levels. Pontusson, Rueda & Way (2002) confirm this notion as they found that wage setting institutions, though broadly combating inequality, have a far stronger effect on the lower-end of the distribution. These arguments underscore why wage setting institutions aim to bolster the bargaining power of workers. As an imbalance of bargaining power may either result in lower wages or lower rates of investment and thus suboptimal outcomes (Bental & Demougin, 2010). Ultimately, the notion that wage setting institutions have a stronger effect on the lower end of the distribution supports the argument that they effectively are ‘protecting the workers’. The ILO claims that “removing ‘rigidities’ in the labour market will not bring about full

employment; it will only worsen workers’ well-being, increase inequality and insecurity, and promote industries that compete on cheap labour rather than technological innovation” (Berg,

9 and bottom end of the distribution, as these drive on abstract human skills or interaction, such as caretakers and managers. In contrast, middle-income earners are the primary target for technology-based substitution, resulting wage polarisation (Goos, et al., 2009). In a world of technological change, developing countries may also face increased difficulty in developing their labour markets, as technological development mainly replaces the routine jobs in manufacturing, which is a crucial sector for economic development (Chenerey, 1986). This inhibits development since labourers in low-income countries now need to adapt to a relatively higher skill level, increasing the threshold for entering the manufacturing sector. Consequently, leaving technological development unregulated imposes a negative effect on both labour market efficiency and equity. However, if a country is characterised by a rigid labour market, the magnitude of the shock caused by skill-biased technological growth is mitigated, as the direct effect on wages is smaller (Fortin & Lemieux, 1997).

2.5 The labour markets of developing countries: imperfect & informal

In the discussion concerning developing countries, the same caveats are relevant. Labour market institutions differ widely between countries and said institutions are also expected to have a varying effect on labour market – and socio-economic – outcomes. In contrast to the market distortion hypothesis, pushed by the neo-classical model of the labour market, Freeman (2009) found that labour market institutions do not necessarily inhibit development. Rather, it is found that, in several countries, such institutions even stimulate development. For example, they were found to be an essential cause for China’s growth (Meng, 2000; Knight & Song, 2005). Further, in Argentina, labour market institutions were found to preserve social stability in terms of crisis (Freeman, 2009). The point that labour market institutions do not necessarily inhibit development is underscored by the argument that both developed and developing countries constitute the top and bottom performers in terms of labour market rigidity (Forteza & Rama, 2001). This evidence hints that the aforementioned caveats are relevant for developing countries as well.

10 Freeman (2009) argues that informality is modelled improperly in conventional literature on developing country labour markets. He argues that the informal sector is not just a last choice safety net, as some workers and firms actually prefer the informal sector (Freeman, 2009). In a case study on Mexico, Hanson (2010) argued that labourers preferred to work in the informal sector since they would have to contribute significantly less to the country’s social security system. Such findings indicate that a rigid labour market may in fact reduce labour market efficiency in labour markets. Furthermore, due to sizeable informal sectors, the insider-outsider effect of labour unions is larger. More specifically, unions have faced difficulties in integrating informal workers under their wing, which diminishes their efficacy over the labour market at large (Birchall, 1999). This effect is expected to surge when the size of informal sector grows larger, since a larger group of outsiders makes it more difficult to enter the formal market for workers on the one hand, and more difficult for unions to operate on the other. Research on wage differentials between the formal and informal sector points out that in some instances self-employed workers in the informal sector earn as much or even more than their formal counterparts (Bosch, et al., 2007). Thus, workers may be incentivised to stay in the informal sector.

Aside from wage setting institutions themselves being heterogenous, they also have divergent effects on labour market outcomes between various groups. Given that there still is a considerable gender gap in labour force participation – especially in developing countries – the effects of wage setting institutions are expected to be unequal (ILO, 2019). Similarly, it has already been established that wage setting institutions affect certain groups in varying ways (Neumark & Wascher, 2004). Betcherman (2014) underscores this argument by suggesting that the low skilled, the young and women are less likely to be covered by such institutions. Therefore, one can expect that the efficiency raising component may be positive for some groups, while negative or non-existent for the other. Ultimately, imperfect labour market should theoretically benefit from stronger wage setting institutions. However, it has been proven that the insider-outsider effect in combination with poor governance of wage setting institutions makes it considerably difficult for said institutions to develop their competence in developing countries.

11

2.6 Explanation of hypotheses

It has become evident from the literature review that wage setting institutions are subject to a wider debate in the field of development economics. Nevertheless, the main proposition in this thesis is that labour markets in developing countries are inherently imperfect, primarily due to a substantial informal sector (Fields, 2011). Wage setting institutions function as instruments to combat these market imperfections and must therefore a priori have a positive effect on labour market outcomes in developing countries. The literature review presented a far bleaker picture, as the imperfections in the labour markets of developing countries impede the functioning of wage setting institutions. Wage setting institutions function perfectly as the core focal point for several reasons. First, they have a direct influence on the price of labour and therefore can be assumed to be, at the very least, correlated with labour market outcomes. Second, these type of wage setting institutions are predominantly debated in academia in the analysis concerning the regulation of labour markets. In academia, wage setting institutions are often analysed in the light of labour market rigidities. For instance, Blau & Kahn (1996) posit that the process of wage determination is subject to a larger degree of coordination – be that from the government or labour unions. As a result from this, labour unions are also classified as wage setting institutions, as they consolidate their bargaining power in order to ‘protect the workers’ (Cazes, 2002). Therefore, in this thesis, the concept of wage setting institutions will cover labour unions and their activity, but also the minimum wage; since all such concept are either directly or indirectly linked to the wage setting process in a labour market.

Following the reasoning that wage setting institutions directly influence the price of labour, one can assume that they are on net making the labour market more equitable. Berg (2015) argued that removing labour market institutions for the goal of flexibility and competitiveness only reduces workers’ well-being and increases income inequality. Furthermore, she claims that a liberalised labour market will result in zero-sum competition. This indicates that wage setting institutions across the board are combatting inequities. However, Piketty (2014) claims that the scope of wage setting institutions is merely effective at the bottom end of the distribution and as such only partially combats the issue of inequality. Nevertheless, the claim of the ILO that they ‘protect the workers’ thus holds merit. In the literature review it became evident that once wage setting institutions gain a stronger position in the market, i.e. cover a larger group of workers, they have considerably more bargaining power. Following this line of reasoning, the first hypothesis is formulated as follows:

Hypothesis 1: In developing countries with stronger wage setting institutions, the income distribution is expected to be more egalitarian.

12 of labour market institutions as they inherently make the market less competitive (Nickell & Layard, 1999). Therefore, the overarching story thus remains that stronger wage setting institutions are associated with a lower labour market participation rate. Consequently, the following hypothesis is formulated:

Hypothesis 2: In developing countries with strong wage setting institutions, the labour market is expected to have a lower labour market participation rate.

In the remainder of this thesis, these two hypotheses are empirically analysed. One important aspect of this analysis resides in the operationalisation of the concepts of interest. Throughout the analysis, equity will be proxied by income inequality, measured in terms of the Gini coefficient. The latter provides an approximation of the distribution of wages, without suffering from a substantial loss of observations. Efficiency will be operationalised by the labour market participation rate of formal employment as a percentage of total population aged 15 or over.

4. METHODOLOGY

4.1 Identification

Given that this thesis’ focal point centres around the core concepts of efficiency and equity, this section will exemplify that these concepts are treated by two distinct empirical models. This section concentrates on the empirical strategy for this thesis. Any discussion with regards to the sources of the data and any potential alterations thereof are discussed in section 4.2. Nevertheless, it is important to highlight that the analysis will be conducted using three different datasets, which require slightly different empirical specifications. The first will be a cross-section of the largest possible number of developing countries, whereas the other two are panel data with an emphasis on either developing or developed economies.

In the first model, the hypotheses are tested in two cross-section regression analyses. More specifically, in equation (1), the Gini coefficient will be regressed on wage setting institution and a set of control variables. In equation (2), the right-hand side will stay the same, but the labour market participation rate will be the main variable of interest. These two model specifications are expressed as follows:

𝐺𝑖𝑛𝑖_𝑖 = 𝛼 + 𝛽1𝑈𝐷𝑒𝑛𝑠_𝑖 + 𝛽2𝑀𝑖𝑛𝑊𝑎𝑔𝑒_𝑖 + 𝛽3𝑊𝑎𝑔𝑒𝐶𝑜_𝑖 + 𝛽3𝑈𝑃𝑜𝑤𝑒𝑟_𝑖+ Σ𝛽_𝑖X_𝑖+ 𝜀 𝐿𝑎𝑏𝑃𝑎𝑟𝑡_𝑖 = 𝛼 + 𝛽1𝑈𝐷𝑒𝑛𝑠_𝑖 + 𝛽2𝑀𝑖𝑛𝑊𝑎𝑔𝑒_𝑖 + 𝛽3𝑊𝑎𝑔𝑒𝐶𝑜_𝑖 + 𝛽3𝑈𝑃𝑜𝑤𝑒𝑟_𝑖+ Σ𝛽_𝑖X_𝑖+ 𝜀 Where the dependent variables Gini and LabPart are the Gini coefficient and the labour market participation rate for country i. 𝛼 is a constant, UDens reflects the rate of labour union density,

MinWage is the minimum wage of each respective country in US$. WageCo represents a score

– from 1 to 10 – that reflects how wages are determined in a country, a high score reflects that wages are determined at the firm level, whereas a low score reflects that wages are set nationally. UPower is an index that ranges from 0 to 1, which represents the judicial power of labour unions, in which a higher index implies relatively more power for labour unions. Finally, X𝑖 are a set of controls and 𝜀 is an error term.

In the second analysis, the specifications are similar to equations (1) and (2) specified above, but adjusted to incorporate the time dimension, as the dataset under scrutiny consists of panel

13 data. The period of time for the analysis ranges from 2000 up until 2016. Moreover, it uses a similar set of controls as above, with some minor exceptions.

The third analysis focuses on the developed world only. In the light of the focus of this thesis, it functions as a control group in order to find whether the efficacy of wage setting institutions may change once developing economies evolve into developed economies. This analysis uses, similar to the prior analyses, two different specifications that focus on regressing the Gini coefficient and the labour market participation rate on the explanatory variables. Given the wider availability of data, more variables could be included in the analysis. For this reason, both specifications are more comprehensive. Furthermore, this analysis is based on panel data, with a time dimension for the years 1990 up until 2017. The two empirical specifications are expressed as follows:

𝐺𝑖𝑛𝑖_𝑖𝑡 = 𝛼 + 𝛽1𝑈𝐷𝑒𝑛𝑠_𝑖𝑡+ 𝛽2𝑀𝑖𝑛𝑊𝑎𝑔𝑒_𝑖𝑡+ 𝛽3𝐶𝑜𝑙𝑙𝐵𝑎𝑟_𝑖𝑡 + 𝛽4𝑊𝑎𝑔𝑒𝐶𝑜_𝑖𝑡+ 𝛽5𝑈𝑛𝑃𝑜𝑤_𝑖𝑡 + 𝛽6𝑆𝑡𝑟𝑖𝑘𝑒𝑠_𝑖𝑡+ Σ𝛽_𝑖𝑡X_𝑖𝑡+ 𝜀

𝐿𝑎𝑏𝑃𝑎𝑟𝑡𝑖𝑡 = 𝛼 + 𝛽1𝑈𝐷𝑒𝑛𝑠𝑖𝑡+ 𝛽2𝑀𝑖𝑛𝑊𝑎𝑔𝑒𝑖𝑡+ 𝛽3𝐶𝑜𝑙𝑙𝐵𝑎𝑟𝑖𝑡 + 𝛽4𝑊𝑎𝑔𝑒𝐶𝑜𝑖𝑡

+ 𝛽5𝑈𝑛𝑃𝑜𝑤_𝑖𝑡+ 𝛽6𝑆𝑡𝑟𝑖𝑘𝑒𝑠_𝑖𝑡+ Σ𝛽_𝑖𝑡X_𝑖𝑡+ 𝜀

Where CollBar represents the collective bargaining coverage rate for country i in year t, Strikes reflects the number of strikes in country i in year t. All the other variables, including the dependent, explanatory and control variables are specified similar to the prior specifications. The first set of control variables include a variety of variables that capture other types of labour market institutions, such as employment protection, hiring and firing regulations, social security-related indicators and active labour market policies. The classification for these other types is based on Ochel (2005) and Botero et al. (2003) in the sense that the overarching concepts applied for these variables are under the dimension of either employment protection or social security – and active labour market policies if available. Additionally, a control variable for the size of the informal sector is included for the analyses on developing countries. This variable is derived from Medina & Schneider (2018). A more detailed explanation regarding the measurement of these variables will be given in section 4.2.

The second set of controls includes a wide array of variables that are potential determinants of variation in income inequality and the labour market participation rate. One major determinant of institutions – and their effectiveness – in general is whether the legal system is based on common law or civil law. Whereas the former is largely based on precedents, the latter is highly statutory in nature, which implies that institutions are far more codified and formalised in the latter (Botero, et al., 2003). Thus, a dummy indicating the source for the legal system will be included as a control. Furthermore, given that institutions are one of the fundamentals for economic development, as per North and collaborators (North, 1981; North & Thomas, 1973; Jones, 1981), including a measure for economic development must be relevant. For this reason, a measure for GDP per capita – adjusted for PPP – is included in the analysis. Additionally, following Bradley & Stephens (2012), a measure for trade openness is included, which will be represented by an overarching score given to countries ‘freedom to trade internationally’ as formulated by the Fraser Institute. This score ranges from 1 (autarky) to 10 (completely open

(3)

14 economy). The intuition behind this inclusion is that the labour market will be more dynamic in an open environment and may thus require more stringent regulations or may be subject to more efficient market dynamics. Trade openness will also be covered by the percentage of exports and imports of GDP in the third analysis, due to data constraints. Another aggregate measure which may influence the relationships as proposed by the model is the extent of government expenditure. Intuitively, the measures classified under labour market policy and especially social security are costly. Thus, one can assume that the more a government spends relatively, the stronger the effects of labour market institutions are on labour market equity and efficiency (Bradley & Stephens, 2012). Therefore, a measure that indicates government spending is included, which is defined as general government final consumption expenditure as a percentage of GDP.

4.2 Data

As has been established, this analysis will cover three different datasets that all have varying scopes regarding the overarching narrative in this thesis. The main difference between these datasets resides in the countries sampled. The first focuses broadly on what the International Monetary Fund classifies as developing countries6, resulting in a total sample of 148 countries (IMF, 2018). This sample is to be found in Appendix A. Some countries, such as Cuba, North Korea and Somalia are excluded by the IMF due to a lack of sufficient or reliable data and are therefore also missing in the dataset.

The first dataset, a cross-section on developing countries, functions as the main test of the role of wage setting institutions for the developing world. The sample consists of a wide set of developing countries, ranging from emerging markets (e.g. China, Brazil and Poland) to underdeveloped nations (e.g. Chad, Nicaragua, Myanmar). Due to lacking data availability over time, a cross-sectional analysis is necessary. In Appendix B, an overview of the coverage and sources of data for this particular dataset is presented. Due to data availability, most of the observations in this dataset are based on a calculated average value for the period 2010 up until 2015. The exceptions to this are the variables Union Power, Employment Law and Social

Security. These are derived from Botero et al. (2003) and explain the relative differences

between legal systems of countries based on the assumption that these are rigid in nature. Furthermore, the data on active labour market policies is from a more recent year, 2016, but as these seem to only marginally change over the years it justifies its inclusion in the dataset. The second dataset, which is constituted of panel data for a set of developing countries7, functions as a test for the time-invariance of the efficacy of wage setting institutions as countries develop. One caveat to this proposition is that the sample included in this dataset is confined to different rules than the prior cross-section. More specifically, with regards to selection, all countries that were regarded as developing by the International Monetary Fund up until 2001 are considered for this dataset. Subsequently, the final sample was simply determined by data

6 _{The IMF classifies countries as developing based on the (1) per capita income level (2) export diversification and (3) the} degree of integration into the global financial system (IMF, 2019).

15 availability. As a result, it must be underscored that this dataset suffers considerably from selection bias. Consequently, data on African countries is the least prevalent and thus, Africa is underrepresented in this dataset. The final sample consists of 25 countries and has data for 17 years, 2000 up until 2016. The countries in this sample are characterised as being a step further in the development process, ranging from post-Soviet economies in Central and Eastern Europe to Brazil, Turkey, South Africa and China. In appendix C, an overview of the sources and coverage of the data is presented.

The third dataset, which is a panel dataset on developed countries, functions as a control group. It can point out whether the efficacy of wage setting institutions is subject to divergent patterns at different stages of economic development. Similarly, it can point that wage setting institutions, or labour market institutions in the broader sense, have a different function in developing countries. The dataset for this analysis consists of 27 countries – including North America, the EU15, Japan and others – and covers the period of 1990 up until 2017. Despite the conventional better data availability of data for developed countries, this dataset is also suffering from lacking data, as especially the first decade under analysis suffers from some gaps in the data. However, a wider number of indicators were available, which allowed me to measure some of the concepts under scrutiny more precisely. In appendix D, an overview of the sources and coverage of the data is presented.

Overall, as can be inferred from the data overviews, there are multiple issues regarding the availability of data. Even though this does not impede the ability to conduct statistical analysis, it is a significant caveat that has to be kept in mind when interpreting the results. Furthermore, table 1, below, exemplifies an explanation of the main variables used in the analyses.

Table 1: Description of the variables

Abbreviation Variable Description Usage

Gini Gini Coefficient of Income. All LabPart Labour Market Participation rate (in % of the working population of 15 or over). All UDens Union density rate, percentage of union members among employees (either in % or a ratio). All MinWage Minimum Wage (in US$); in exchange rates. All

WageCo

Decentrality of wage determination: 0 if completely centralised, 10 if completely decentralised

(firm-level) All

UPower

Index on the state of collective relations law, both regarding labour union power and collective disputes, a higher index implies more power for labour unions OR A dummy variable indicating a 1 if employers are legally required to bargain with unions and 0 if not.

First indicator: Dataset 1; second indicator: Datasets 2 & 3 CollBar Collective bargaining coverage rate. Dataset 3

Strikes The number of strikes in a country in a year Dataset 3 EmpPro Employment Protection, scale of 1-6 in which 6 represents stricter regulation Dataset 3 HiringFiring Hiring and firing regulations: 0 if completely rigid, 10 if completely flexible All EmpLaw Index on the state of laws concerning employment, a high index implies more regulation Dataset 1 SocSec Index on the state of social security laws, a high index implies a stronger social security system Dataset 1

SocSec2

A dummy variable indicating 1 if the social security system covers the risk of unemployment and 0 if

not Dataset 2

UB Unemployment Benefits (as % of GDP) Dataset 3 ALMP

Active labour market policy in a country, to what extent do Labour market policies help unemployed people to reskill and find new employment (including skills matching, retraining, etc.)? [1 = not at all; 7 = to a great extent] OR Active labour market policy (as % of GDP)

First description: Dataset 1; second description: Dataset 3

16

5. RESULTS

5.1 Descriptive statistics

This chapter will start with a description of the data used for this thesis, by means of providing certain descriptive statistics as well as discussing correlations and some minor graphical analysis. Accordingly, table 2, below provides the descriptive statistics for the forthcoming statistical analyses. Datasets 1 and 2 both focus on developing countries, though the coverage differs significantly between these two samples.8 _{The time period for both panel analyses also}

requires some clarification. In dataset 2, the period 2000-2016 is chosen due to this set including a number of post-Soviet economies. Including the 1990s in the dataset would result in noise throughout the data. Since the set of developed countries is less susceptible to noise, this decade is included in dataset 3. Nevertheless, both panels suffer from the inclusion of the 2008 financial crisis, but this will be taken into consideration while conducting the analyses.

Table 2: Descriptive statistics

Variables Dataset N Mean St.Dev. Min Max Skewness Kurtosis

Gini 1 114 40.042 8.068 24.69 69.4 0.694 3.625 2 325 37.202 8.999 23.72 64.76 0.673 3.076 3 593 34.104 6.062 21.6 53.9 0.712 3.144 LabPart 1 128 61.560 12.413 32.8 89.5 -0.040 2.516 2 425 59.629 7.027 45.056 77.217 0.456 3.161 3 756 61.176 5.815 47.333 78.367 -0.099 3.494 UDens 1 65 18.235 13.115 2 63 1.038 3.958 2 279 21.279 12.194 3 73.618 0.918 3.809 3 660 35.310 22.041 7.578 92.991 0.989 2.885 MinWage 1 121 179.811 171.803 0 915 1.834 7.228 2 356 256.632 192.005 5 1047 1.292 5.497 3 300 1239.637 522.348 239.476 2721 0.467 2.425 WageCo 1 119 6.453 1.122 0 8.593 -2.141 10.288 2 423 6.593 1.461 2.02 8.851 -1.059 3.863 3 514 5.796 1.826 1.833 9.5 -0.049 1.886 UPower 1 54 0.444 0.126 0.1875 0.711 -1.880 2.409 2 408 0.792 0.407 0 1 -1.436 3.063 3 700 0.520 0.500 0 1 -0.080 1.006 CollBar 3 538 63.058 26.29 11.3 98.5 -0.578 2.067 Strikes 3 589 233.481 380.233 0 2768 2.612 11.327 HiringFiring 1 120 4.690 1.060 0 6.692 -1.063 5.717 2 423 4.356 1.203 1.336 8.802 -0.005 2.622 3 513 4.690 1.849 1.333 8.825 0.499 2.061 EmpLaw 1 54 0.475 0.182 0.148 0.828 0.181 2.161 EmpPro 3 547 2.115 0.861 0.257 4.833 0.342 4.002 SocSec 1 54 0.476 0.242 0 0.850 -0.194 1.758 SocSec2 2 408 0.750 0.434 0 1 -1.155 2.333 UB 3 612 1.084 0.764 0.07 4.71 1.329 5.382 ALMP 1 100 2.983 0.791 1.54 4.96 0.307 2.500 3 572 0.698 0.46 0.08 2.7 1.220 4.804 1 121 32.142 10.010 12.11 65.1 0.427 3.348

8 _{See Appendices A, B and C for clarification} Open

Economic Openness: 0 if completely closed, 10 if completely open for trade OR Trade openness, proxied by taking the sum of imports and exports as a % of GDP

First description: Datasets 1 & 2; second description: Dataset 3 LegalOrg Legal Origins of a country's judiciary system (0 if civic law, 1 if common law and 2 if Islamic law) All

17 InfSec 2 400 27.500 9.813 10.47 54.17 0.403 2.534 GDPc 1 146 11576.89 15393.91 752.6819 127091.8 4.290 27.694 2 425 15341.94 7400.28 1790.172 35230.52 0.293 2.318 3 756 33960.83 15029.16 8273.406 107640.6 1.567 6.792 Open 1 122 6.748 1.005 3.333 8.562 -0.748 3.602 2 425 7.343 0.842 3.6 9.139 -0.922 5.411 3 756 99.520 86.907 16.014 442.62 2.226 7.360 LegOrg 1 147 0.429 0.597 0 2 1.053 3.093 2 425 0.160 0.367 0 1 1.855 4.440 3 756 0.333 0.472 0 1 0.707 1.500 GvmtSpend 1 132 15.755 5.969 5.178 49.764 1.916 10.763 2 425 16.030 3.631 7.069 22.924 -0.307 1.838 3 752 18.797 4.478 6.837 27.935 -0.474 2.774

After carefully analysing and reviewing the data, certain variables provide some interesting insights on how the trajectory of a country’s labour market institutions and outcomes evolves at different stages of economic development. With regards to this, the distributions of the dependent variables, efficiency and equity are particularly interesting. More specifically, the labour market participation rate seems to be similar in all three datasets. However, the lower standard deviation and less extreme minimum and maximum for the third dataset shows that at higher levels of development, the labour market participation rate is relatively more stable. In terms of equity, the Gini-coefficient was found to be lower for developed countries, indicating that developed countries tend to be more equal in terms of income. Furthermore, as the mean value for dataset 2 is also lower than that of dataset 1, it can be argued that once a county starts developing, its income distribution is expected to become more equal.

Another interesting insight is that, overall, wage setting institutions tend to be stronger in the developed world, as union densities and coordination of wages are all found to have higher scores for dataset 3. One exception to this role is the variable indicating the power for labour unions. A possible explanation for this is that dataset is characterised by a large number of post-Soviet economies, that along traditional socialist lines, have relatively powerful labour unions. Though it may also indicate that in more competitive, developed labour markets, unions are expected to weaken (Addison, et al., 2015).

The descriptive statistics show that all variables tend to stabilise with development, indicating developed countries are more similar. This statement is, however, a bold one. Given that dataset 1 is a cross-section on a wide set of a developing countries, it is considerably more subject to between-country heterogeneity. For this reason, it must be underscored that the mean, minimum and maximum values for this dataset only tells a part of the overarching story regarding the efficacy of labour market institutions. Furthermore, regarding causal interference, some variables require to be transformed. As becomes evident from analysing the skewness and kurtosis, some variables are far from normally distributed and would thus benefit from a logarithmic transformation. Such a transformation is conducted for these variables, i.e. MinWage, GDPc and GvmtSpend in dataset 1 and Strike, UB, GDPc and Open in dataset 3.

5.2 Correlations and graphical analysis

18 developing countries9. The two main concepts of interest, the Gini and labour market participation are found to be significantly positively correlated. This entails that a country with a relatively high level of labour market participation generally also has a relatively unequal income distribution. Furthermore, the Gini is found to be negatively related to union density, which implies that a country with a higher union density is associated with a more equal income distribution. Interestingly, WageCo is negatively associated with the Gini, implying that more central determination of wages has the opposite effect on the income distribution. Labour market participation, on the other hand, was found to only be negatively associated with the minimum wage. This indicates that the hypotheses regarding the efficacy of wage setting institutions are aligned with divergent patterns in equity and efficiency of developing labour markets. Furthermore, the table in appendix E shows that WageCo is significantly correlated to HiringFiring, which indicates that multicollinearity may pose an issue in the analysis.

Table 4 below presents the correlation coefficients for the panel data on developing countries. Once again, the dependent variables are positively correlated. This implies that the association of a relatively high labour market participation rate with a less egalitarian income distribution holds for both datasets on developing countries. The Gini is found to be correlated to all explanatory and control variables with the exception of union density. Moreover, the coefficient seems to imply that stronger wage setting institutions are associated with a more equal income distribution. An exception to this is the negative coefficient of WageCo, which suggests the opposite. Labour market participation is similarly correlated to all explanatory variables and controls. Again, union density is the exception here, as no significant correlation is found. All explanatory variables display a negative sign, implying that stronger wage setting institutions are associated with a lower labour market participation rate. One caveat is that with regards to multicollinearity, the minimum wage is strongly associated (r = 0.80) with GDP per capita.

Table 4: Correlation coefficients for dataset 2. *** p<0.01, ** p<0.05, * p<0.1

Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (1) Gini 1 (2) LabPart 0.313*** 1 (3) UDens -0.051 -0.038 1 (4) MinWage -0.305*** -0.255*** -0.202*** 1 (5) WageCo -0.405*** -0.201*** -0.318*** -0.135** 1 (6) UPower -0.361*** -0.197*** -0.103* 0.291*** 0.121** 1 (7) HiringFiring -0.241*** 0.159*** -0.098 -0.362*** 0.504*** -0.009 1

9 _{Table 3 only displays the correlation coefficients for the dependent and explanatory variables. For a table including the} coefficients of the control variables, one may refer to appendix E.

Gini LabPart MinWage (log) UDens WageCo UPower Gini 1 LabPart 0.236** 1 MinWage (log) -0.005 -0.392*** 1 UDens -0.220** -0.100 -0.052 1 WageCo -0.233** 0.048 -0.153 -0.016 1 UPower -0.062 -0.041 0.105 0.193 -0.113 1

19 (8) SocSec -0.287*** -0.320*** 0.440*** 0.145*** 0.054 -0.059 -0.205*** 1 (9) InfSec 0.276*** 0.238*** 0.117* -0.494*** -0.196*** -0.082 0.194*** -0.490*** 1 (10) GDPc -0.335*** -0.221*** -0.302*** 0.796*** 0.149*** 0.164*** -0.139*** 0.271*** -0.589*** 1 (11) Open -0.269*** -0.164*** -0.384*** 0.241*** 0.369*** 0.172*** -0.025 0.170*** -0.466*** 0.405*** 1 (12) LegalOrg 0.321*** 0.389*** -0.223*** -0.217*** -0.261*** -0.046 0.077 -0.516*** 0.381*** -0.239*** -0.189*** 1 (13) GvmtSpend -0.456*** -0.300*** 0.132** 0.363*** 0.132*** 0.272*** -0.048 0.606*** -0.312*** 0.434*** 0.244*** -0.211*** 1

Table 5 below10 _{presents the correlation coefficients for the panel dataset of developed}

countries. Glancing through this table and its extended equivalent highlights several issues regarding multicollinearity. More specifically, WageCo and CollBar are strongly associated to each other (r = -763), which implies that if a country has a relatively large coverage of its collective bargaining agreements, it also has a relatively centralised process of wage determination. Similarly, the Minimum Wage is strongly associated with a country’s GDP per capita and employment protection seems to be strongly associated with legal origins. Even though this may not cause problems for the forthcoming analysis, it must still be handled carefully. Shifting our focus to the variables at hand, it becomes evident that the two dependent variables are now not associated. This implies that for developed countries, a relatively more efficient labour market does not entail a more – or less – unequal distribution of income. However, the Gini is associated with all explanatory variables with half of them exhibiting the expected sign, and the other half, i.e. WageCo, UPower and Strike exhibiting that stronger wage setting institutions are associated with a more unequal income distribution. The picture for labour market participation is somewhat similar, as it is also significantly correlated to a majority of the explanatory variables, excluding the minimum wage.

Table 5: Correlation coefficients for dataset 3. *** p<0.01, ** p<0.05, * p<0.1

The influence of wage setting institutions on labour market outcomes seems to become more pronounced when development increases, which indicates that the efficacy wage setting institutions may indeed evolve during economic development. On the other hand though, its effect also becomes more ambiguous at higher stages of economic development; which is represented by the less uniform signs of the correlation coefficients for the panel dataset on developed countries. Ultimately, it seems as if labour market outcomes – i.e. the distribution of income and the labour market participation rate – may indeed follow a U-shape distribution. In order to analyse whether this may hold some truth, graphical analysis will be conducted. Furthermore, such analysis may help in finding out whether there are significant outliers to be observed. With regards to this, three scatterplots will be scrutinised (4 in the case of efficiency).

10 _{Table 5 only represents the correlation coefficients for the dependent and explanatory variables. For a table that includes all} variables, one may refer to appendix F.

20 Each plot represents a different step in the process of economic development, ranging from developing to the developed world.11

Figure 2: Scatterplots of the Gini-coefficient against economic development for datasets 1, 2 and 3. In the figure, the fitted values lines are plotted as well.

Figure 3: Scatterplots of the labour market participation rate against economic development for dataset 1. The left one includes the outliers, whereas the right one excludes them. In the figure, the fitted value lines are plotted as well.

Figure 4: Scatterplots of the labour market participation rate against economic development for dataset 2 and 3. In the figure, the fitted value lines are plotted as well.

21 The distribution of equity does not exhibit any tendency towards a U-shape distribution. Rather, it seems as if economic development is negatively associated with the Gini coefficient, which implies that once a country develops it is found to have a more egalitarian income distribution. Nevertheless, the magnitude of the effect seems to be generally stronger for countries at a relatively advanced stage of development; hence the more downward-sloping trendlines for the latter two graphs. The graphs for labour market participation indicate that the U-shape distribution may indeed hold merit for efficiency, since the developing countries exhibit a negative relationship between GDP per capita and efficiency, whereas in the developed world, a more developed country is expected to have a more efficient labour market. Furthermore, Figure 4A presents that for developing countries, the spread of the labour market participation rate is smaller for larger values of GDP per capita. This implies that one’s starting position is relevant when assessing the efficacy of wage setting institutions.

5.3 Empirical results

Table 6 shows the results of the cross-sectional analysis on developing countries. The table presents six specifications, all derived from the equations in chapter 4. The first and fourth are baseline specifications focusing on one of the two dependent variables. The second and fifth are the full model specifications. Finally, the third and sixth specifications include a dummy for the poorest countries in the sample which identifies whether the inference would differ if controlled for countries with a GDP per capita lower than the mean12. The choice of explanatory variables in the model specifications is derived from the correlation coefficients and follows from a stepwise regression analysis. In terms of the explanatory power of the specifications, the model seems to perform better in predicting equity than efficiency, as both the R-squared and its adjusted equivalent are consistently higher for the specifications for equity. The specifications including the low-income dummy seem to have the most explanatory power, although the differences are not substantial. For efficiency, none of the explanatory variables are significant, regardless of the specification. It must however be noted that excluding one of the indicators for wage setting institutions, the coefficient for the minimum wage reveals a significantly negative effect.13 Finally, White tests have been conducted in order to investigate whether heteroskedasticity poses an issue for the analysis. For all specifications, I fail to reject the null hypothesis that heteroskedasticity prevails in the data. Thus, no robust standard errors are used.

Table 6: Empirical results of the cross-section analysis on developing countries. Standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1

Gini Labour market participation Variables Specification 1 Specification 2 Specification 3 Specification 4 Specification 5 Specification 6

Indicators for wage setting institutions

Union density -0.145 1.920 1.464 6.195 (0.106) (8.992) (8.943) (12.52)

Minimum wage (log) 0.673 -2.684 0.131 -1.675 (1.935) (2.290) (2.075) (1.844) Decentrality of wage determination -2.295** -1.998** -2.041** -0.0500

(1.089) (0.918) (0.918) (1.289)

12 _{In these models, the log of GDP per capita is disregarded due to multicollinearity.}

22

Labour union power -9.398 -12.49 (11.72) (13.88)

Control variables

Employment laws -3.444 -3.410 (7.655) (7.664)

Hiring and firing regulations -0.401 -0.211 (1.637) (1.692) Social security -7.032 -6.859 -6.665 -14.71* (7.527) (7.512) (8.361) (8.223) Active labour market policy -7.042*** -6.338***

(1.786) (1.706)

Informal sector -0.325** -0.320** 0.248 0.284* (0.143) (0.144) (0.151) (0.154) GDP per capita (log) 2.540 -3.119

(2.011) (2.450)

Trade openness 2.419* 2.416* -0.876 -1.725 (1.273) (1.276) (1.674) (1.659) Legal Origins 7.288* 6.005 -2.051 -2.711 (3.991) (3.906) (3.884) (3.938) Government spending (log) -1.566 -1.455 -7.356 -5.551 (4.782) (4.777) (5.441) (5.397)

Low income (dummy) -3.224 -3.254

(2.616) (3.917) Constant 59.04*** 51.82** 74.43*** 80.26*** 113.7*** 99.10*** (13.45) (20.57) (15.74) (15.92) (24.98) (21.05) Observations 38 39 39 38 49 49 R-squared 0.186 0.543 0.542 0.085 0.300 0.283 F-statistic 1.89 3.33*** 3.32*** 0.76 2.14* 1.98*

23 low income countries, the extent of social security laws also becomes negatively related to labour market efficiency at the 10% level. Overall though, hypothesis 2 is to be rejected for this cross-section of developing countries. There is not enough statistical evidence to conclude that stronger wage setting institutions result in lower labour market participation.

In order to test whether the efficacy of wage setting institutions alters when a developing country progresses economically, the following panel data analysis is conducted. In tables 7 and 8 below, the outcomes of the model specifications are exhibited. Prior to delving into the outcomes, it is insightful to discuss the econometric setup of the model. Using a Breusch-Pagan test, it became evident that for both equity and efficiency, OLS is not a suitable method of analysis. Rather, as shown by a Hausman test, both equity and efficiency are best suited to be analysed using a random effects model. The only exception to this is specification 1 in table 7, which is estimated using a fixed effects model. These methodologies are consistent for all specifications in the tables. This implies that for equity, all time-invariant variables, i.e. labour union power, social security and legal origins, are omitted from the results. Similar to the prior analysis, stepwise regression is used. Of the specifications presented below, the second one is the conventional test of the empirical specification. Specifications 3 and 4 represent a full test of the model while incorporating the effects of the financial crisis of 2008-2009. More specifically, specification 3 uses year fixed effects, which could identify whether the income inequality was substantially different from the trend in individual years. Specification 4 on the other hand simply discards the crisis years from the analysis.

Table 7: Empirical results for the panel analysis on the Gini for developing countries. (Robust) standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

Variables Specification 1 Specification 2 Specification 3 Specification 4

Indicators for wage setting institutions

Union density -0.0957* -0.0842 -0.320*** -0.0821 (0.0469) (0.0555) (0.0505) (0.0655) Minimum wage -0.00452* -0.00311 -0.00734** -0.00282 (0.00258) (0.00335) (0.00362) (0.00345) Decentrality of wage determination 0.446 -0.0541 0.344 -0.145

(0.733) (0.556) (0.472) (0.548) Labour union power -6.105 -10.28*** -5.826

(4.918) (1.562) (5.093)

Control variables

Hiring and firing regulations 0.250 -1.408*** 0.224 (0.286) (0.533) (0.270) Social security 3.096 9.890*** 3.877

(7.401) (2.574) (7.704) Informal sector 0.249 0.339*** 0.310* (0.159) (0.0776) (0.172) GDP per capita 9.75e-05 -6.92e-05 0.000115

24

Robust standard errors Yes Yes No Yes Year fixed effects No No Yes No

Crisis dummy No No No Yes

Observations 189 187 187 155

R-squared 0.063 0.325 0.684 0.339 F-statistic 1.50 35.15*** 345.54*** 35.24*** Number of countries 23 23 23 23

Looking at the R-squared and F-statistics of the specifications, one can infer that those which incorporate the financial crisis by including year-fixed effects has more explanatory power. In specification 3, all explanatory variables but the decentrality of wage determination significantly negatively influencing the Gini. This implies that stronger wage setting institutions cause the income distribution to become more egalitarian in emerging market economies. Even though none of the other specifications present significant results for the explanatory variables, it must be added that when using conventional standard errors, union density would also be significantly negative for specification 2. These findings are in line with hypothesis 1. Ultimately, for this panel dataset of developing countries, there is enough statistical evidence to conclude that stronger wage setting institutions result in a more equal distribution of income. Analysing the control variables provides some interesting insights. Trade openness, for instance, was found to be positively significant in specifications 2 and 4, but negatively in specification 3. Furthermore, the size of the informal sector and social security were found to be positively significant, whereas hiring and firing regulations and government spending were found to be negatively influencing the Gini. Evidently, countries that have a larger informal sector and a social security system safeguarding unemployment were found to have a less egalitarian income distribution. Stronger employment protection and greater government spending on the other hand cause the income distribution to equalise. The effect of trade openness is, based on these results, ambiguous. Regarding specification 3, the years surrounding the financial crisis exerted no significantly different values from the trend. Aside from that, no other years were found to have a significant effect.

Table 8: Empirical results for the panel analysis on the labour market participation of developing countries. (Robust) standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

Variables Specification 1 Specification 2 Specification 3 Specification 4

Indicators for wage setting institutions

Union density -0.000706 0.0357 -0.00101 0.0552 (0.0284) (0.0273) (0.0393) (0.0364) Minimum wage 0.000717 -0.00320* -0.00504 -0.00217 (0.00129) (0.00175) (0.00312) (0.00167) Decentrality of wage determination -0.124 0.000914 -0.367 -0.0982

(0.314) (0.308) (0.371) (0.335) Labour union power -3.932 -3.777 -6.919*** -3.501 (3.961) (4.469) (1.172) (4.346)

Control variables

Hiring and firing regulations -0.0650 0.586 -0.0373 (0.207) (0.424) (0.199) Social security -4.248 -3.174 -4.771 (4.471) (2.002) (4.616) Informal sector 0.0422 -0.0736 0.0351 (0.0925) (0.0592) (0.111) GDP per capita 0.000170* 0.000259*** 0.000157*