حداقل دستمزد و فقر خانگی: تعادل عمومی شبیه سازی ماکرو - میکرو برای آفریقای جنوبی

Minimum wages are used widely in developing countries to protect vulnerable workers, reduce wage inequality, and lift the working poor out of poverty. The political popularity of minimum wages stems in part from the fact that the policy offers a means for redistributing income without having to increase government spending or establish formal transfer mechanisms (Card & Krueger, 1995). The challenge to policymakers is to find that wage level that is considered fair given workers’ needs and the cost of living, but does not harm employment or a country’s global competitiveness (Figueiredo and Shaheed, 1995 and Gindling and Terrel, 2010). Economic theory as well as applied work in the area of minimum wages tends to focus on the employment effects of the policy. Fields and Kanbur (2007), however, argue that a more pertinent welfare question is how minimum wages affect poverty. This is especially true in developing countries where poverty reduction remains a major policy goal, and where attachment to the labor market does not necessarily mean not being poor. However, minimum wages are sometimes considered ineffective in targeting the poor in developing countries, mainly because the poor are more likely to be unemployed, self-employed, or employed in informal sectors where enforcement of minimum wage policy is challenging (Lustig & McLeod, 1997). This study undertakes an empirical investigation of how minimum wages impact on poverty in South Africa. We develop an ex ante macro–micro modeling framework that captures both the economywide and micro-level distributional effects of minimum wages. The analysis contributes to earlier ex post and ex ante analyses by modeling the various channels through which minimum wages affect poverty within an integrated framework. South Africa presents a unique case study, not only because minimum wage policy design differs from one country to the next in terms of targeting or the level of the wage minimum in relation to the poverty line, but also because of the unique characteristics of its labor market (Section 3 elaborates). Our results are therefore not necessarily representative of developing countries in general. That said, the conceptualization and methods used are novel and, we believe, widely applicable. The paper is organized as follows. Section 2 discusses pertinent theoretical and modeling issues with reference to the existing literature. Section 3 provides background on South Africa’s labor market, including details of the minimum wage policy. Section 4 introduces the model and Section 5 presents and discusses the simulation results. Section 6 draws conclusions. 2. Theoretical considerations and analytical framework Both ex post and ex ante evaluations of the employment effects of minimum wages examine the factor demand responses (or wage elasticities) at the level of the firm or sector. While these employment responses are a crucial starting point, evaluations of the poverty effects of minimum wages require a more comprehensive understanding of the overall distributional effects of the policy. In an extension to the standard employment-focused minimum wage theory, Fields and Kanbur (2007) show that when poverty impacts are of interest it is no longer just the wage elasticity that matters. Also important are the extent and nature of income sharing within or between households (i.e., a consideration of how employment and wage changes in the labor market are linked back to specific households and individuals within households) and the aversion to poverty as captured by the choice of poverty measure. Fields and Kanbur (2007) make a valuable contribution in highlighting these additional channels. However, their extended model still rests on a number of simplifying assumptions.1 In practical applications where many of the strict assumptions of their model are relaxed, several further poverty impact channels have been identified. These include considerations around compliance with and/or enforcement of minimum wage legislation in different sectors (Gindling & Terrel, 2010), how minimum wages impact on labor supply and wage levels in sectors not covered by minimum wage legislation (Gindling & Terrel, 2010), and the location of minimum wage workers (as potential beneficiaries) within the overall income spectrum (Burkhauser et al., 1996 and Lustig and McLeod, 1997). The impact of minimum wages on production costs and consumer prices may also have welfare implications (Bird & Manning, 2008). Finally, efficiency wage theorists claim higher wages reduce shirking, raise the efficiency of labor utilization, or directly increase worker productivity due to increased nutritional intake (see Shapiro & Stiglitz, 1984). These induced labor productivity gains may reduce employment losses associated with minimum wages. Gindling and Terrel (2010) provide a thorough review of the literature on minimum wages and poverty in developing countries. Among these, the ex post cross-country studies surveyed by them are fairly consistent in the message that an inverse relationship exists between (changes in) poverty and (changes in) the minimum wage, while results from country-level analyses using micro-data, which include ex post and ex ante approaches, are more mixed. They stress that ex post analyses such as their own in Honduras fail to isolate the effects of minimum wages per se from those of other external shocks. This common problem of ex post analyses is solved by ex ante models, which allow for real-time policy analysis in a manner that is less time-consuming and less costly than ex post analyses ( Bourguignon & Spadaro, 2005). However, ex ante models rely on explicit assumptions about economic behavior and as such should be viewed as exploratory tools rather than predictive tools. Somewhat surprisingly, there are few examples of ex ante simulations of minimum wages and poverty. Models developed by Hertz, 2002 and Bird and Manning, 2008 respectively provide for an interesting comparison. Hertz’s (2002) household survey-based model identifies those domestic workers in South Africa that benefit from a hypothetical minimum wage and those that are most likely to lose their jobs (based on econometrically estimated employment probabilities) under a range of wage elasticities. Through combining wage income gains and losses at the household-level he is able to demonstrate how poverty declines as long as the wage elasticity parameter is inelastic. Bird and Manning (2008) simulate the poverty impact of minimum wages in Indonesia using a short-run model in which firms mitigate the effect of increased production costs by raising consumer prices rather than by shedding jobs. While about one-fifth of households benefit from higher wages, the remainder requires a higher level of income to acquire the same basket of goods as before. The authors conclude that while minimum wages are mildly progressive under the targeting and the funding mechanisms assumed, they are unlikely to significantly reduce poverty in Indonesia. The above two ex ante approaches capture partial effects only, with two major omissions: first, the models restrict firms’ cost mitigation options; and, second, neither model considers the indirect implications of price changes in the economy. In reality firms are likely to use a combination of lay-offs, charging higher prices, and profit reduction to mitigate production cost increases. Poverty impacts are likely to be sensitive to the mitigation strategy followed. For example, when the wage elasticity is low and the employment response is small, more minimum wage workers will benefit, but minimum wage sectors will be forced to raise prices. Conversely, when the wage elasticity is high, fewer workers will benefit from higher wages but price increases will be smaller. The indirect effects of minimum wages stem from the fact that higher prices in minimum wage sectors spill over into other sectors via inter-industry linkages. Ultimately, aggregate demand and consumption patterns will be affected by price and household income changes, which in turn have secondary implications for production and employment in the economy via various demand channels. We therefore argue that economywide general equilibrium models are better equipped than partial equilibrium models to deal with the complex cost mitigation strategies and indirect price and demand effects associated with minimum wage policies. Yet, despite their discernable advantages, not many general equilibrium applications of minimum wages and poverty exist. Standard neoclassical modeling applications include analyses by Holland, Bhattacharjee, and Stodick (2006) in the United States and Paes de Barros, Corseuil, and Cury (2001) in Brazil, while Gibson and Van Seventer (2000) apply a structuralist general equilibrium model to evaluate the effect of wage increases in South Africa. The limited number of general equilibrium modeling applications probably relates to the fact that traditionally these models are not used in the analysis of microeconomic policies. Even modern neoclassical computable general equilibrium (CGE) models which are often highly disaggregated are not particularly well suited to analyzing poverty and distributional changes at the level of the household. This stems from the restrictive assumption that income distributions within representative household and factor groups in the model are static (Bourguignon & Spadaro, 2005). For poverty assessments, at least, survey-based partial equilibrium models (or microsimulation models) in the mold of those described above are more appropriate. However, by linking macro- (CGE) and microsimulation models sequentially, the advantages of each modeling approach can be exploited. Such an integrated modeling framework is developed here for South Africa and used to evaluate the poverty effects of minimum wages.

Minimum wages are used widely in developing countries to protect vulnerable workers, reduce wage inequality, and lift the working poor out of poverty. The political popularity of minimum wages stems in part from the fact that the policy offers a means for redistributing income without having to increase government spending or establish formal transfer mechanisms (Card & Krueger, 1995). The challenge to policymakers is to find that wage level that is considered fair given workers’ needs and the cost of living, but does not harm employment or a country’s global competitiveness (Figueiredo and Shaheed, 1995 and Gindling and Terrel, 2010). Economic theory as well as applied work in the area of minimum wages tends to focus on the employment effects of the policy. Fields and Kanbur (2007), however, argue that a more pertinent welfare question is how minimum wages affect poverty. This is especially true in developing countries where poverty reduction remains a major policy goal, and where attachment to the labor market does not necessarily mean not being poor. However, minimum wages are sometimes considered ineffective in targeting the poor in developing countries, mainly because the poor are more likely to be unemployed, self-employed, or employed in informal sectors where enforcement of minimum wage policy is challenging (Lustig & McLeod, 1997). This study undertakes an empirical investigation of how minimum wages impact on poverty in South Africa. We develop an ex ante macro–micro modeling framework that captures both the economywide and micro-level distributional effects of minimum wages. The analysis contributes to earlier ex post and ex ante analyses by modeling the various channels through which minimum wages affect poverty within an integrated framework. South Africa presents a unique case study, not only because minimum wage policy design differs from one country to the next in terms of targeting or the level of the wage minimum in relation to the poverty line, but also because of the unique characteristics of its labor market (Section 3 elaborates). Our results are therefore not necessarily representative of developing countries in general. That said, the conceptualization and methods used are novel and, we believe, widely applicable. The paper is organized as follows. Section 2 discusses pertinent theoretical and modeling issues with reference to the existing literature. Section 3 provides background on South Africa’s labor market, including details of the minimum wage policy. Section 4 introduces the model and Section 5 presents and discusses the simulation results. Section 6 draws conclusions. 2. Theoretical considerations and analytical framework Both ex post and ex ante evaluations of the employment effects of minimum wages examine the factor demand responses (or wage elasticities) at the level of the firm or sector. While these employment responses are a crucial starting point, evaluations of the poverty effects of minimum wages require a more comprehensive understanding of the overall distributional effects of the policy. In an extension to the standard employment-focused minimum wage theory, Fields and Kanbur (2007) show that when poverty impacts are of interest it is no longer just the wage elasticity that matters. Also important are the extent and nature of income sharing within or between households (i.e., a consideration of how employment and wage changes in the labor market are linked back to specific households and individuals within households) and the aversion to poverty as captured by the choice of poverty measure. Fields and Kanbur (2007) make a valuable contribution in highlighting these additional channels. However, their extended model still rests on a number of simplifying assumptions.1 In practical applications where many of the strict assumptions of their model are relaxed, several further poverty impact channels have been identified. These include considerations around compliance with and/or enforcement of minimum wage legislation in different sectors (Gindling & Terrel, 2010), how minimum wages impact on labor supply and wage levels in sectors not covered by minimum wage legislation (Gindling & Terrel, 2010), and the location of minimum wage workers (as potential beneficiaries) within the overall income spectrum (Burkhauser et al., 1996 and Lustig and McLeod, 1997). The impact of minimum wages on production costs and consumer prices may also have welfare implications (Bird & Manning, 2008). Finally, efficiency wage theorists claim higher wages reduce shirking, raise the efficiency of labor utilization, or directly increase worker productivity due to increased nutritional intake (see Shapiro & Stiglitz, 1984). These induced labor productivity gains may reduce employment losses associated with minimum wages. Gindling and Terrel (2010) provide a thorough review of the literature on minimum wages and poverty in developing countries. Among these, the ex post cross-country studies surveyed by them are fairly consistent in the message that an inverse relationship exists between (changes in) poverty and (changes in) the minimum wage, while results from country-level analyses using micro-data, which include ex post and ex ante approaches, are more mixed. They stress that ex post analyses such as their own in Honduras fail to isolate the effects of minimum wages per se from those of other external shocks. This common problem of ex post analyses is solved by ex ante models, which allow for real-time policy analysis in a manner that is less time-consuming and less costly than ex post analyses ( Bourguignon & Spadaro, 2005). However, ex ante models rely on explicit assumptions about economic behavior and as such should be viewed as exploratory tools rather than predictive tools. Somewhat surprisingly, there are few examples of ex ante simulations of minimum wages and poverty. Models developed by Hertz, 2002 and Bird and Manning, 2008 respectively provide for an interesting comparison. Hertz’s (2002) household survey-based model identifies those domestic workers in South Africa that benefit from a hypothetical minimum wage and those that are most likely to lose their jobs (based on econometrically estimated employment probabilities) under a range of wage elasticities. Through combining wage income gains and losses at the household-level he is able to demonstrate how poverty declines as long as the wage elasticity parameter is inelastic. Bird and Manning (2008) simulate the poverty impact of minimum wages in Indonesia using a short-run model in which firms mitigate the effect of increased production costs by raising consumer prices rather than by shedding jobs. While about one-fifth of households benefit from higher wages, the remainder requires a higher level of income to acquire the same basket of goods as before. The authors conclude that while minimum wages are mildly progressive under the targeting and the funding mechanisms assumed, they are unlikely to significantly reduce poverty in Indonesia. The above two ex ante approaches capture partial effects only, with two major omissions: first, the models restrict firms’ cost mitigation options; and, second, neither model considers the indirect implications of price changes in the economy. In reality firms are likely to use a combination of lay-offs, charging higher prices, and profit reduction to mitigate production cost increases. Poverty impacts are likely to be sensitive to the mitigation strategy followed. For example, when the wage elasticity is low and the employment response is small, more minimum wage workers will benefit, but minimum wage sectors will be forced to raise prices. Conversely, when the wage elasticity is high, fewer workers will benefit from higher wages but price increases will be smaller. The indirect effects of minimum wages stem from the fact that higher prices in minimum wage sectors spill over into other sectors via inter-industry linkages. Ultimately, aggregate demand and consumption patterns will be affected by price and household income changes, which in turn have secondary implications for production and employment in the economy via various demand channels. We therefore argue that economywide general equilibrium models are better equipped than partial equilibrium models to deal with the complex cost mitigation strategies and indirect price and demand effects associated with minimum wage policies. Yet, despite their discernable advantages, not many general equilibrium applications of minimum wages and poverty exist. Standard neoclassical modeling applications include analyses by Holland, Bhattacharjee, and Stodick (2006) in the United States and Paes de Barros, Corseuil, and Cury (2001) in Brazil, while Gibson and Van Seventer (2000) apply a structuralist general equilibrium model to evaluate the effect of wage increases in South Africa. The limited number of general equilibrium modeling applications probably relates to the fact that traditionally these models are not used in the analysis of microeconomic policies. Even modern neoclassical computable general equilibrium (CGE) models which are often highly disaggregated are not particularly well suited to analyzing poverty and distributional changes at the level of the household. This stems from the restrictive assumption that income distributions within representative household and factor groups in the model are static (Bourguignon & Spadaro, 2005). For poverty assessments, at least, survey-based partial equilibrium models (or microsimulation models) in the mold of those described above are more appropriate. However, by linking macro- (CGE) and microsimulation models sequentially, the advantages of each modeling approach can be exploited. Such an integrated modeling framework is developed here for South Africa and used to evaluate the poverty effects of minimum wages.

In this study an ex ante modeling framework was used to evaluate various employment, production, income, and poverty effects of the set of minimum wages that were introduced in South Africa during 2002–06. The CGE-microsimulation framework developed for this purpose is shown to be a useful advance over the type of partial or general equilibrium ex ante models that have been used in the analysis of minimum wages and poverty. Whereas the CGE component of the model properly accounts for the important price and indirect demand effects of the policy that are often ignored or understated in partial equilibrium models, the microsimulation component assesses the distribution of employment gains and losses and the implications for household income at the individual unit of observation. Results presented show that, ceteris paribus, out of a labor force of 11 million in 2000, minimum wages may force anywhere from 350,000–500,000 people into unemployment. Ex post we now know that the economy in fact created over two million jobs in the first few years since 2000 ( Casale et al., 2004); hence one interpretation of these results is that job creation could have been up to 25% higher in the absence of minimum wages. Also, many of these jobs were created in informal sectors or were set up as short-term contracts, which is perhaps indicative of an attempt by employers to circumvent minimum wage legislation. Our results show that the majority of expected job losses are among domestic workers, whereas labor force statistics show no decline in domestic employment from 2000 to 2010. This may be due to wage restructuring, strong growth in demand for domestic services due to economic growth that has benefited middle- to high-income households in particular, and a high degree of non-compliance with the minimum wage laws. The main purpose of this study, however, was to move beyond these employment effects and to understand the poverty implications of minimum wages at different wage elasticity levels. Adequately assessing these poverty implications requires the linking of this employment picture into a comprehensive household and economywide framework. The heart of the paper is an attempt to do this and to show that the resulting insights for policy justify the effort. Results suggest that for poverty lines ranging from R2 000 (indicating ultra-poverty) to R4 000 (normal poverty) per capita per annum and a moderate wage elasticity (η = 0.7) that constitute the benchmark scenario in the paper, the minimum wages implemented in South Africa post-2000 will reduce the poverty headcount rate (P0) by about 3–5%, ceteris paribus. At lower poverty lines the decline in poverty is much smaller, while the depth of poverty (P1) increases substantially. Thus, while good for poverty generally, minimum wages raise inequality among the poor—by 24.9% in the benchmark scenario. These adverse distributional effects relate, firstly, to the fact that job losses are likely to be biased against the very poor. The beneficiaries of the policy are therefore more likely to be those among the sub-minimum wage workers that are initially better off or non-poor, suggesting that the policy is not well targeted for poverty alleviation. Secondly, those among the poor that do gain from the policy gain at the expense of over two-thirds of the poor that see their disposable income levels decline due to job losses, declining returns to factors of production, and rising prices. The model is clear that skilled workers and wealthy households shoulder by far the largest share of the cost burden since they account for a large proportion of overall consumption. However, in addition to these better-off sections of the population, those poor that fail to benefit from minimum wages, such as the unemployed or those not directly targeted by the policy, are especially vulnerable to any resultant price increases. Indeed, rising unemployment, rising prices, and rising inequality among the poor, more so than declines in real income of wealthy households, remain important reasons why minimum wages are such controversial policy instruments. This study has shed some light on the issue of minimum wages and poverty in South Africa through the use of a more comprehensive framework of analysis than is commonly used in such analyses. While South Africa is rather unique among developing countries, thus limiting the extent to which our results should be generalized for developing countries, the analytical method employed here can be adapted to different countries fairly easily. The fact that the net impact of minimum wages on poverty in South Africa is sensitive to the additional macro–micro linkages that we highlight makes a case that this is worth doing. With respect to South Africa’s experience with minimum wages, some questions remain to be answered. First, our analysis could be usefully complemented by ex post assessments such as that by Dinkelman and Ranchhod (2010) of the effects of minimum wages on employment and, importantly, wage distributions. For domestic workers in South Africa, Dinkelman and Ranchhod (2010) find considerable bunching of the wage distribution from above and below around the minimum wage and no evidence of employment losses. Whereas a bunching up of the lower tail of the wage distribution at the level of the minimum wage would be indicative of compliance by firms (as investigated explicitly by Bhorat et al. (2010)), it may also happen that wage increases at the upper end of the distribution are curtailed over time in an effort to cross-subsidize the minimum wage. Ex post analyses may also reveal other cost mitigation strategies by firms, such as reductions in profits, which were not considered here. These mitigation strategies may ultimately have different poverty implications compared to those that involve price hikes and lay-offs. Second, our analysis, which has focused on the immediate implications of factor market substitution and price shocks, did not consider the longer-term dynamic effects associated with changing investment behavior. Such an analysis would require a more careful consideration of how current investment flows are converted into future capital stocks, or indeed, as Gibson and Van Seventer (2000) show, how changes in government fiscal and monetary policies may shape the outcome of minimum wages.