Employment transitions and the cycle of income inequality in postindustrial societies

Abstract

This article presents the claim that the current transition from service to knowledge employment impacts income inequality in a manner that is comparable to the previous transition from agriculture to industry. This contention is tested by an employment transition index that calculates the difference of employment between the higher- and lower-paid sectors of the economy. The calculation of this index is consistent with Kuznets’s view that income inequality increases during the early stages of industrialization due to the presence of a small and higher-wage modern sector that encroaches on the total numbers employed in the large and lower-wage traditional sector. However, income inequality eventually declines with continued industrialization as more workers enter the modern sector of the economy. Results confirm the central argument of this study for a panel of 25 Organisation for Economic Co-operation and Development countries from 1980 to 2008. According to the findings, classical Kuznetsian variables such as sector-dualism are not significant and/or signed in unanticipated directions in their prediction of income inequality. Instead, the employment transition index returns robust negative associations with income inequality and consistently outperforms sector-dualism. Furthermore, knowledge employment returns positive and significant connections with the dependent variable net of the employment transition index. These results confirm that both between- and within-sector employment patterns are key determinants of income inequality.

Keywords

Employment transitions income inequality knowledge economy knowledge employment Kuznets inverted-U hypothesis labor-force trends work and occupations

Introduction

In the literature on income inequality, few theoretical propositions are more controversial than the ideas of Simon Kuznets (1955). According to Kuznets, income inequality increases during the early stages of industrialization as a small segment of the labor-force is employed in the industrial sector. However, these lofty levels of inequality eventually decrease as workers continue to migrate to the higher-wage segments of the national economy. Perhaps most controversially, based on these observations Kuznets would famously assert, albeit with some personal skepticism and doubt, that income inequality would decline into the foreseeable future with further economic development and industrialization. From these ideas grew an early body of evidence that returned strong support for the aptly named Kuznets Inverted-U hypothesis (Adelman and Morris, 1978; Ahluwalia, 1976; Chenery and Syrquin, 1975; Williamson and Lindert, 1980). Yet, by the 1980s, an equally large body of literature directly contradicted Kuznets’s contentions as income inequality began to increase in advanced industrial societies (Bluestone, 1990; Bluestone and Harrison, 1982; Harrison and Bluestone, 1988; Levy and Michel, 1991; Levy and Murnane, 1992).

Over the years, many observe that the rise of income inequality in postindustrial societies is driven by the changing returns to skill in the new economy. Referred to as skilled-biased technological change, this argument is based on the premise that the computerization of work enhances the demand for non-routine cognitive skills thereby increasing the wages of educated employees (Acemoglu and Autor, 2011; Autor et al., 2003; Autor and Dorn, 2010; Black and Spitz-Oener, 2010; Katz, 2000; Krueger, 1993; Levy and Murnane, 1992). Similarly, the empirical research shows that inequality is no longer driven by the transfer of labor from agriculture to industry, but by the wide variation of earnings between skilled and unskilled workers in what is traditionally referred to as the service sector (Chevan and Stokes, 2000; Kim and Sakamoto, 2008; Lee et al., 2011; Moller et al., 2009; Nelson and Lorence, 1988).

In light of these observations, this article attempts to amend the Kuznets Inverted-U hypothesis to account for the rise of the knowledge economy. While in agreement with the Kuznetsian view that inequality rises and falls as workers migrate from lower- to higher-paid forms of employment, this study’s point of departure is its rejection of the assumption that a majority of the labor-force will eventually be employed in the higher-paid industrial sector. This emphasis on the agricultural–industrial transition is the precise limitation that leads Kuznets to claim that inequality will continue to decline in the future. In contrast, this investigation adopts the view that economic development is characterized by the presence of numerous and coexisting employment transitions that generate a continual cycle of income inequality (Korzeniewicz and Moran, 2009; Moran, 2005). Simply put, just as the transition from agriculture to industry generated a rise and fall of inequality, so too is the current transition from labor-intensive service to knowledge employment resulting in a comparable dynamic.

The proposed cycle of inequality hypothesis is tested by an employment transition index (ETI) that captures the impact of between-sector employment shifts on income inequality. This index ranges from 0 to 100 with higher scores denoting a larger gap of employment between the higher- and lower-paid sectors of the economy. This formulation is consistent with Kuznets’s claim that inequality is initially low as a majority are employed in agriculture (larger differential of employment between higher- and lower-paid sectors), rises as workers begin to migrate to manufacturing (smaller differential of employment), and falls with the continued transfer of labor to the industrial sector (larger differential of employment). The proposed index thus contains three major characteristics: first, it retains the Kuznetsian logic of rising and falling inequality with an economy’s transition from lower- to higher-paid employment; second, it accounts for the presence of multiple sectors as opposed to the dual-economy view of Kuznets; and finally, it proportionally weights the agricultural–industrial and service–knowledge transition, respectively, since the distributional impact of these dynamics changes over time in direct proportion with their share of total employment.

This investigation contributes to a well-established literature that attempts to measure various aspects of the Kuznets hypothesis. Some scholars utilize gross domestic product (GDP) per capita and its squared term in a regression analysis to test the argument of lower inequality at higher levels of development (e.g. Lee, 2005; Moller et al., 2009; Nielsen, 1994). Still others employ sector-dualism to explore the association between sectoral wage differentials and income inequality (e.g. Lecaillon et al., 1984; Moller et al., 2009; Nielsen, 1994; Rohrbach, 2009). The latter measure is particularly noteworthy since it is widely used by macro-comparative sociologists as the definitive test of the Kuznetsian framework. However, these approaches do not directly analyze the distributional consequence of between-sector employment trends which is a major, if not central, cause of inequality for Kuznets. To this end, the ETI is the only measure that directly quantifies the labor-force composition component of the Kuznetsian framework. Furthermore, this index extends Kuznets’s arguments to account for both the agricultural–industrial and service–knowledge transition, respectively.

The findings are largely consistent with the arguments of this study as the ETI is a significant predictor of income inequality. However, two features of the new economy provide additional clarity regarding the connection between employment shifts and income inequality. First, knowledge employment produces a robust positive association with the dependent variable net of the ETI. Since the ETI controls for the distributional effect of between-sector employment patterns, this finding shows that within-sector inequality in the knowledge sector is an important determinant of income inequality in postindustrial societies. Second, the inclusion of female labor participation in the regression models substantially reduces the association between the ETI and income inequality. By extension, this shows that wage differentials between men and women may account for a substantial share of the between-sector effect on earnings inequality. The article begins below with a detailed summary of the literature on employment shifts and income inequality.

From the Inverted-U hypothesis to the great U-Turn

Simon Kuznets (1955) was primarily concerned with analyzing the distribution of income in national economies and, more specifically, with finding the connection between economic development and income inequality. Examining a limited amount of data for England, Germany, and the United States, Kuznets saw the beginnings of a possible emergent trend. He discovered that the lofty levels of income inequality found in these nation-states during the early-1900s started to decline in the 1920s. Furthermore, this decline of inequality was unfolding during a time when the aforementioned countries were experiencing rapid economic growth. In this way, the parallel timing of these trends led Kuznets to assert that income inequality expands during the early stages of industrialization only to diminish with further economic development.

Kuznets relies heavily on a dual-economy model of employment demographics to explain the relationship between development and income inequality. According to his model, the early phases of industrialization are characterized by the presence of a small and higher-wage modern sector, i.e., manufacturing, that encroaches on the total numbers employed in the large and lower-wage traditional sector, i.e., agriculture. At first, the migration of workers to industry expands wage differentials given that a smaller proportion of the labor-force is employed in the modern sector vis-à-vis agriculture. But continued industrialization allows more workers to enter the modern sector thereby resulting in a decline of income inequality.

From these origins grew a robust empirical tradition that utilized cross-sectional development and inequality data to present evidence of Kuznets’s theoretical conjecture. On the one hand, some researchers plot economic development along the x-axis and income inequality along the y-axis to show an identifiable Inverted-U pattern of inequality. On the other hand, others perform a regression analysis that returns a negative and significant association between a quadratic term for GDP per capita and income inequality (e.g. Adelman and Morris, 1978; Ahluwalia, 1976; Chenery and Syrquin, 1975; Williamson and Lindert, 1980). Nielsen (1994) further explores the Kuznetsian view of inequality by analyzing the effect of sector-dualism, that is, the wage differential between agriculture and all other sectors in the economy, for a cross-section of 56 countries during 1970. He finds that sector-dualism is a positive and significant predictor of inequality net of relevant control variables (see also Lee et al., 2007; Mahutga et al., 2011).

However, macroeconomic trends during the 1970s and 1980s would foment a literature that raised doubts regarding the Kuznets hypothesis. Coined the Great U-Turn by Harrison and Bluestone (1988), many discovered that the declining levels of inequality in rich countries started to increase in the 1970s (Bluestone, 1990; Levy and Michel, 1991; Levy and Murnane, 1992). Moreover, empirical research was returning evidence that various factors such as declining employment in manufacturing, increased participation of women in the labor-force, higher levels of imports from foreign nation-states, waning unionization rates, and the rising prominence of the financial sector were newly emergent variables in the determinacy of this mounting inequality (Alderson, 1999; Alderson and Nielsen, 2002; Bluestone and Harrison, 1982; Harrison and Bluestone, 1988; Volscho and Kelly, 2012; Western and Rosenfeld, 2011).

A number of works also showed evidence that directly contradicted the views of Kuznets. To reiterate, earlier tests of the Inverted-U hypothesis would include a quadratic term for GDP per capita in a regression analysis to test the Kuznetsian proposition of lower inequality at higher levels of development (e.g. Ahluwalia, 1976). Similarly, contemporary works also find evidence of decreasing inequality at higher levels of GDP per capita when performing a panel analysis of countries from all income levels (Barro, 2000; Lee, 2005; Lee et al., 2007; Tsai, 1995). However, the results differ when restricting the analysis to postindustrial societies, as newly available longitudinal data return a robust positive association between a quadratic of GDP per capita and income inequality (Alderson and Nielsen, 2002; Deininger and Squire, 1996; Moller et al., 2009).

A particularly interesting study in this regard comes from Nielsen and Alderson (1997), who examine the determinants of inequality for a panel of U.S. counties during the years 1970, 1980, and 1990, respectively. They find that sector-dualism is strongly linked with inequality in 1970, only, while this relationship fails to produce a robust association in 1980 and 1990. These authors expand the regional scope of their analysis to study the determinants of inequality in 16 Organisation for Economic Co-operation and Development (OECD) countries for the years 1967–1992. They find that sector-dualism shares a precarious relationship with income inequality and is outperformed by various Great U-Turn variables (Alderson and Nielsen, 2002). All-in-all and as stated by Nielsen and Alderson (1997), ‘[t]he declining importance of some of the factors traditionally associated with the impact of industrial development on income inequality … suggest that the determination of income inequality in advanced industrial societies is the outcome of a new set of processes’ (p. 30).

The service sector and income inequality

Early empirical studies relied heavily on cross-sectional data to show the Kuznets Inverted-U pattern of development and income inequality (e.g. Barro, 2000; Nielsen, 1994; Tsai, 1995).¹ But scholars started to look beyond the Inverted-U hypothesis as newly available longitudinal data returned evidence that invalidated Kuznets’s claims. It is in this context that many turned to an assessment of the service sector.² Particularly important in this regard is the purported bifurcated nature of the service sector’s occupational structure which contains higher-wage jobs such finance, insurance, and real-estate (FIRE), on the one hand, as well as lower-wage jobs such as retail, sales, and personal service occupations, on the other (Fernandez-Macias, 2012; Levy and Murnane, 1992; Lorence and Nelson, 1993; Sassen, 1996, 1998).

According to Nelson and Lorence (1988), various subsets of service occupations augment inequality differently for men and women. Inequality for men stems from the higher earnings of males in professional and managerial positions while inequality for women results from the lower wages earned by females in retail and low-skilled services. Chevan and Stokes (2000) examine the determinants of inequality across a panel of 784 U.S. metropolitan regions. They find that higher-wage service jobs such as business, professional, and FIRE occupations are significantly linked to inequality from 1980 to 1990 but not significantly associated during 1970 to 1980. Moller et al. (2009) examine the effect of FIRE jobs on U.S. county-level inequality from 1970 to 2000. They discover that FIRE occupations significantly augment inequality while all other non-FIRE service occupations do not produce a significant effect.

In this way, there exists a large literature on the importance of the service sector for the recent growth of income inequality (see also Lee et al., 2011). However, while FIRE’s positive link to income inequality is interpreted as proof of within-sector inequality, this evidence may merely point to an inadequate conceptualization of the services. In other words, it is possible that the literature’s reliance on the classical definition of the service sector, which contains a broad mix of occupations with little similarity, may inflate its within-sector effect.

Distinguishing between knowledge and labor-intensive service employment

Given the literature, it is clear that the growing prominence of the service sector is an important determinant of recent income inequality trends. But to produce significant results, many scholars engage in a strategy of subdividing the services into higher- and lower-paid occupations. However, the popularity of this approach reveals the problems that are associated with defining the services in a way that truncates all non-agricultural and non-industrial forms of employment into a single category. It is clear that there is a need to disaggregate the occupations in what is traditionally referred to as the service sector. A good starting point to achieve this objective is the literature on the knowledge society.

For more than half a century, advanced industrial societies have experienced a general shift from a goods-producing to a service-producing economy. Theoretical works argue that rich countries are evolving into a knowledge-based society wherein production is increasingly reliant on the usage of intellectual capabilities rather than physical labor or natural resources (Drucker, 1969; Machlup, 1962; Stanback, 1979). To a certain extent, knowledge and innovation was always a central determinant of economic development: from railroads and the industrial revolution, during the 1800s, to scientific management and the assembly line, during the 1900s. In this way, the creation of new technology and/or designs that increase the scalability and idiosyncrasy of production is not a novel phenomenon. However, what differentiates the new economy from previous socio-economic forms is the establishment of a knowledge sector that institutionalizes knowledge creation and innovation as a perpetual feature of postindustrial societies (Drucker, 1993).

It is important to note that the formation of the knowledge sector is intrinsically linked to ‘globalization’ and the increasing returns to education. Specifically, the corporate community engaged in a concerted effort to remedy the problem of stagnant corporate profits during the 1970s. An important part of this strategy was the large-scale liquidation of fixed corporate assets resulting in a flood of financial investments into cutting-edge industries from artificial intelligence, computing, biotechnology, information technology, nanotechnology, and robotics (Davis, 2009; Schumpeter, 1939). In addition, corporate executives simultaneously engaged in cost-cutting strategies such as the outsourcing of production, streamlining organization structures, and the creation of labor-saving technologies (Kalleberg, 2011). Together these strategies increased the demand for college-educated employees due to, first, the nature of work in many knowledge-intensive occupations and, second, the heightened managerial intensity required by the lean and mean model of corporate governance (Goldstein, 2012; Gordon, 1996).

The role of education as outlined above is similar to the explanation provided by the skill-biased technological change literature. According to this view, the computerization of work enhances the demand for non-routine cognitive skills thereby increasing the wages of educated employees (Acemoglu and Autor, 2011; Autor and Dorn, 2010; Autor et al., 2003; Black and Spitz-Oener, 2010; Katz, 2000; Krueger, 1993; Levy and Murnane, 1992). However, increased demand for cognitive skills is not solely the result of computerization but is also due to the expansion of the knowledge sector. Furthermore, skill-biased technological change fails to recognize that the demand for cognitive skills stems from the corporate community’s cost-cutting agenda which frequently occurs at the expense of unionized rank-and-file employees (Fligstein and Shin, 2007; Kristal, 2013).

Over the years, many attempted to identify and measure what is often called knowledge workers, creative employees, and an expert-managerial class (e.g. Florida, 2002; Howkins, 2001; Wright, 1997). According to Florida (2002), labor-intensive service jobs are made up of low-wage and low-autonomy occupations with examples ranging from ‘food-service workers, janitors and grounds keepers, personal care attendants, secretaries and clerical workers, and security guards and other service occupations’ (p. 71). As for knowledge employment, this author distinguishes between what he calls a creative core and creative professionals. The former is defined by the production of novel designs and inventions that are transferable and widely useful such as ‘designing a product that can be widely made, sold, and used; coming up with a theorem or strategy that can be applied in many cases; or composing music that can be performed again and again’ (Florida, 2002: 69). The latter occupations often exercise a great deal of judgment and draws on complex bodies of knowledge ‘to solve specific problems … or combine standard approaches in unique ways to fit the situation’ (Florida, 2002: 69).

In this way, there exists a relatively robust theoretical literature on the knowledge society and a burgeoning field of research that measures the size of the knowledge sector. Nevertheless, there is a surprising shortage of research that tests the aggregate effect of knowledge occupations on income inequality (for an exception, see Rohrbach, 2009). Instead, researchers rely on employment data for individual occupations – that is, science, technology, finance, and business – in what is traditionally referred to as the services, to assess their impact on the national distribution of income (Chevan and Stokes, 2000; Kim and Sakamoto, 2008; Lee et al., 2011; Moller et al., 2009; Nelson and Lorence, 1988). Problematic is that an examination of various subsets of occupations fails to offer an overall assessment of the knowledge sector’s distributional effect.

In closing, what becomes clear from the empirical literature is the relative paucity of research that directly tests the aggregate effect of knowledge employment on income inequality. The next section provides a theoretical framework with which to study employment trends and distributional outcomes. Namely, it attempts to amend the general Kuznetsian framework to take into account the changing nature of employment during the rise of the knowledge society.

The cycle of income inequality: Reframing employment transitions during the rise of the knowledge society

Two features of the Kuznetsian framework are useful to analyze at this point in the discussion. First, the Inverted-U hypothesis revolves around the idea that sectoral shifts in the labor-force shape distributional trends. This is a key contribution of Kuznets as transitions from lower- to higher-paid forms of employment should stimulate an initial rise and eventual fall of income inequality. Second, Kuznets takes these arguments a step further to provide the bold assertion that inequality will continue to decline in the future with economic development. It is clear today that this proposition is not reflected in the empirical reality of rich countries. But it is important to recognize that Kuznets, like most scholars, was heavily influenced by the popular views of his time. One such influential view during the mid-1900s was that industrialization is synonymous with development and represents the final stage of a society’s economic evolution. Given this widespread assumption during the time of his writing, the ‘logical’ conclusion available to Kuznets was that income inequality will continue to decline, as a vast majority of workers will eventually be absorbed into the higher-paid industrial sector.

Regardless, the inaccuracies of Kuznets’s predictions do not imply that employment transitions are irrelevant for income inequality. As already discussed, one of Kuznets’s lasting contributions is his generalized description of distributional patterns during an economy’s transition from lower- to higher-paid forms of employment. Thus, it is possible to reframe this basic logic within a theoretical framework that allows for the existence of multiple and/or overlapping employment transitions as opposed to the stringent dual-economy view provided by Kuznets.³ This reframing allows for the formulation of a cycle of inequality hypothesis which maintains that economic development generates a continual cycle of income inequality that is driven by a constant series of shifts in the sectoral demographics of employment.

In this proposed view, economic development is characterized by a constant series of employment transitions that generate a cycle of income inequality. These employment transitions are driven by advancements in technology and production which causes workers to continually relocate in newly emergent cutting-edge industries. Employment transitions also entail a concurrent shift in wage allocations that generate a rise and fall of income inequality. This point is similar to the Kuznets hypothesis as the shift from lower- to higher-wage occupations is seen here as a major driver of inequality. However, the cycle of inequality hypothesis rejects the notion of a final stage of economic development during which a majority of workers will be employed in a single sector. In this way, knowledge employment does not represent a final phase of economic development as the cycle of inequality should continue in the decades to come.

Finally, the cycle of inequality hypothesis also adopts the view that employment transitions generate a continual drive toward inequality (Korzeniewicz and Moran, 2005, 2009; Moran, 2005). This drive toward inequality is due to, first, the increasingly complex division of labor in newly emergent sectors and, second, the accumulation and concentration of capital via rent-seeking activities. The latter point fits well into a larger literature which shows that financial activities display a cyclical and linear trend that peaks during the founding of new cutting-edge industries (Piketty, 2014; Schumpeter, 1939).⁴ Similarly, others find that product cycles drive boom and bust trends in various economic outcomes and activities such as growth, production, prices, wages, finance, trade, and unemployment (Chase-Dunn, 1998; Freeman and Soete, 1997; Misra and Boswell, 1997; Modelski and Thompson, 1996; Perez, 2003; Wallerstein, 1984).

Measuring Kuznetsian employment shifts: The employment transition index

The discussion above provides a useful framework for understanding between-sector shifts of employment and their relationship to income inequality. Figure 1 shows a stylized illustration of this study’s cycle of inequality hypothesis with three trends of relevance for the distribution of national income.⁵ First, the left-hand side depicts the agricultural–industrial transition that occurred in rich countries from the mid-1800s to the mid-1900s. Observable here is the classical Inverted-U pattern of rising inequality, that is, the broken line, during the nascent stages of industrial development and the eventual decline of inequality with the maturation of this process. This is the precise trend that Kuznets saw in the data during the mid-1900s. However, it is important to note that this timing is a generalization that only applies to Western countries, since other rich nation-states, such as the East Asian NICs, experienced these transitions at a much later point in time.

Figure 1.

Employment transitions and the cycle of income inequality.

Second, the top of Figure 1 shows the general shift from a goods- to service-producing economy. This shift to a service-producing economy is what, in turn, stimulated a renewed upsurge of income inequality in postindustrial societies. It is well known that the mid- to late-20th century is defined by the migration of labor out of industry toward the labor-intensive service and knowledge sector, but equally drastic during this time is the outflow of workers from agriculture. Figure 2 shows employment statistics for OECD countries during the years 1970 to 2008. According to the data, the percentage of workers in industry declined from 35.5 percent in 1970 to 30.1 percent by 2008. However, employment in agriculture fell even more drastically during this period from 16.0 to 3.6 percent. What these patterns show is the general migration of labor away from both agriculture and industry into the service-producing segments of the economy.⁶

Figure 2.

Employment transitions in OECD countries.

This shift from a goods- to service-producing economy produces major implications for the Inverted-U hypothesis. Kuznets was certainly correct to claim that inequality will rise and fall with the agricultural–industrial transition. Nevertheless, what he failed to anticipate is the diminishing importance of these sectors for income inequality given the net migration of workers toward service-producing occupations. According to Figure 2, already by the 1970s, there were an equal number of workers employed in the goods- and service-producing sectors. But by the 2000s, there are approximately two times more workers in service-producing occupations. In this context, it is unlikely that the classical Kuznetsian argument will hold the same predictive power.

Third, the right-hand side of Figure 1 illustrates in more detail the sectors associated with the rise of the service-producing economy and its connection to income inequality. This portion of the model shows that inequality grows in conjunction with the shift from lower-wage services toward higher-wage knowledge. As shown in Figure 2, employment in the labor-intensive service sector is on the decline as it accounts for 33.5 percent of total employment in the year 1970 but only 26.9 percent in 2008. Instead, all of the recent employment gains in advanced industrial countries are concentrated in the knowledge sector as its 14.9 percent share of employment in 1970 increases to 39.4 percent by 2008.⁷ Thus, it is possible that the elevated levels of inequality in postindustrial societies are at least partially driven by the nascent service–knowledge transition.⁸

In light of these observations, this study formulates an employment transition index (ETI) that is designed to measure the impact of between-sector employment patterns on income inequality. This measure is constructed to do the following: first, it captures the argument of rising and falling inequality with an economy’s transition from lower- to higher-paid forms of employment; second, it accounts for the recurrent and overlapping nature of employment transitions as opposed to the dual-economy approach; and finally, it properly weights the overall effect of the agricultural–industrial and service–knowledge transition, respectively, given that the distributional impact of these dynamics changes over time in direct proportion to their share of total employment. In this way, the ETI is measured by the formula

ETI = (G_{d} \times G_{t}) + (S_{d} \times S_{t})

where G_d = |A – I|, G_t = A + I, S_d = |L − K|, S_t = L + K, and A represents employment in agriculture, I employment in industry, L employment in labor-intensive services, and K employment in the knowledge sector, all reported as a percent of the total labor-force.

In this formula, the absolute values are used to measure the difference of employment between the higher- and lower-paid sectors. This is consistent with Kuznets’s argument that inequality is low when a majority of the labor-force is employed in agriculture (larger differential of employment between the higher- and lower-paid sectors), rises as workers begin entering into manufacturing (smaller differential of employment), and falls as more of the labor-force joins the industrial sector (larger differential of employment). Less relevant in this regard is whether the higher- or lower-paid sector maintains a larger share of total employment. Instead, it is the gap of employment between these sectors that is the major element of inquiry. Moreover, proportionate weights capture the declining relevance of the agricultural–industrial transition as these sectors’ share of employment decreases. Conversely, the service–knowledge transition’s impact should increase over time with these sectors’ enlarged share of total employment (see Figure 2). Logistically, the ETI is designed to range from 0 to 100, with a lower score denoting a labor-force that is relatively dispersed between all sectors of the economy, and a higher score signifying more concentration within a single sector. As such, higher ETI values should be negatively associated with income inequality. This is consistent with the Kuznets Inverted-U hypothesis that income inequality rises as industrial employment encroaches on the total numbers employed in agriculture, but that inequality eventually declines as more workers flow into the industrial sector.

Table 1 attempts to further clarify the logistics of the ETI by providing U.S. sectoral employment data from 1980 to 2008. According to these statistics, agriculture accounts for 2.7 percent of the labor-force in 1980 while industry accounts for 31.7 percent during the same year. But by the year 2008, agricultural employment dropped to 0.7 percent and industrial employment fell to 21.1 percent. Particularly noteworthy here is that the difference of employment between agriculture and industry is on the decline and, by extension, is resulting in a decrease of the ETI during the period in question. Furthermore, while these goods-producing sectors account for 34.4 percent of total U.S. employment in 1980, they only make up 22.4 percent in 2008. This shows that the agricultural–industrial transition’s weight in the ETI formula diminishes over time since employment in these sectors is on the decline. As for the service-producing sectors, employment in these segments of the economy is on the rise as service grows from 38.2 to 41.2 percent while knowledge swells from 27.3 to 36.3 percent. Thus, the service–knowledge transition is also contributing to a reduction of the ETI as the difference of employment between these sectors is on the decline. In addition, the weight of the service–knowledge transition in the determinacy of the ETI is increasing over time since these sectors’ share of total U.S. employment expands during the period in question.

Table 1.

U.S. Sectoral Employment Statistics and Employment Transition Index (ETI).

Year	Agriculture	Industry	Goods-Producing	Service	Knowledge	Service-producing	ETI
1980	2.76%	31.67%	34.43%	38.27%	27.30%	65.57%	17.15
1985	3.24%	28.14%	31.38%	41.45%	27.16%	68.62%	17.62
1990	2.90%	26.78%	29.69%	41.30%	29.02%	70.31%	15.73
1995	2.92%	25.29%	28.21%	40.38%	31.41%	71.79%	12.76
2000	2.51%	24.56%	27.07%	39.45%	33.48%	72.93%	10.32
2005	0.69%	22.87%	23.56%	41.70%	34.75%	76.44%	10.54
2008	0.68%	21.75%	22.43%	41.27%	36.30%	77.57%	8.59

Source: ILO at http://laborsta.ilo.org/.

Figure 3 displays a scatter plot of the ETI and Gini coefficient for this study’s sample of 25 OECD countries from 1980 to 2008. There is an undeniable amount of variation around the linear regression line which is reinforced by the relatively low 14.8 percent of the variation in income inequality that is explained by the ETI.⁹ This large unexplained variation may be due to the fact that inequality is conditioned by national and regional socio-political variables (Beckfield, 2006; Fernandez-Macias, 2012; Huber et al., 2006; Lee, 2005; Mann and Riley, 2007). Nevertheless, there is a clear negative association between these covariates.

Figure 3.

Scatterplot of Gini coefficient and employment transition index.

Data

The dependent variable is the pre-transfer Gini coefficient of inequality from the Standardized World Income Inequality Database (SWIID) available at http://myweb.uiowa.edu/fsolt/swiid/swiid.html. The decision to use pre- as opposed to post-transfer inequality is based on the logic that the demographics of employment are more likely to produce a direct effect on market-based inequality.¹⁰ A major methodological issue with cross-national calculations of income inequality is the relative lack of comparable estimates. The incompatibility of data results from differences between countries in their data collection methodologies and accounting practices. Specifically, differences in definitions of earnings (income versus expenditures), population coverage (metropolitan versus all), and units (household versus individual), all contribute to the relative lack of compatible estimates of national income distributions. However, the SWIID optimizes the between-country comparability of the popular UN-WIDER dataset by calculating Gini ratios generated by paring observations with similar income definitions and units to the Luxemburg Income Study (for more detail, see Solt, 2009).

The major independent variable is the ETI. The data required to construct this measure are from the International Labour Organization (ILO) available at http://laborsta.ilo.org/. The ILO calculates the number of workers employed in specific groups of occupations as defined in the International Standard Classification of Occupations (ISCO). The current study aggregates these occupations into categories of sectors to measure the effect of sectoral employment shifts on income inequality. Appendix 1 shows the ISCO occupations and their placement within this study’s conceptualization of the agricultural, industrial, labor-intensive service, and knowledge sector. Unfortunately, the ILO’s occupational scheme varies slightly over time due to the periodic updating of their coding methodology. Thus, differences between the various iterations of the ISCO are such that a perfect transfer of occupations between versions is not possible. Even so, most occupational groups are similar enough to allow the sectors of this study to remain relatively coherent from one version to another.¹¹ In addition, since the ETI accounts for between-sector effects on income inequality, the inclusion of knowledge employment in the regression models is used to explore within-sector effects. Knowledge employment is measured as the percent of labor in the knowledge sector vis-à-vis the total labor-force; this information is from the aforementioned ILO database.

There are three groups of control variables which, unless otherwise noted, are from the World Development Indicators (World Bank, 2013). The first group of covariates represents what many refer to as the internal development model (Nielsen, 1994). Agricultural employment calculates the percentage of the labor-force employed in agriculture and should produce a negative association with income inequality. This variable is converted to its natural log to adjust for positive skew, and these data are from the ILO. Sector-dualism is the absolute value of the percentage of labor employed in agriculture minus agriculture’s contribution to GDP. This measure is also converted to its natural log and should return a positive relationship with the dependent variable. As discussed in earlier sections, sector-dualism is designed to calculate the income gap between agriculture and all other sectors of the economy. The problem with this measure is that it, on the one hand, fails to capture the employment shift component of the Kuznets Inverted-U hypothesis and, on the other hand, does not account for the changing patterns of employment with the rise of the knowledge society. The natural rate of population growth, or the crude birth rate minus the crude death rate, is expressed as a ratio of 1000 and should be positively associated with inequality.

The second cluster of variables represents five popular covariates from the inequality literature. Unemployment rate is the percentage of labor without work but both available and seeking employment. Unemployment is converted into its natural log to adjust for positive skew and should be positively linked with inequality. Tertiary school enrollment is the percent of those attending post-secondary educational institutions vis-à-vis the total age-relevant population. This measure is preferred to secondary school enrollment given the augmented wage premiums garnered by those with college degrees in advanced industrial countries (Autor et al., 2003). Female labor-force participation is measured as a percent of the labor-force and is from the OECD database available at http://www.oecd.org/statistics/. Past research shows an inconsistent association between female labor participation and inequality. Union density is the total unionized workforce expressed as a percent of all workers and should produce a negative association. This variable is from the aforementioned OECD database. Credit to the private sector controls financialization’s impact on inequality (Jerzmanowski and Nabar, 2013; Lin and Tomaskovic-Devey, 2013). This is the total capital given to the private sector, in the form of loans or other financial instruments that establish a claim for repayment, as a percentage of GDP. This variable is transformed into its natural log and should be positively correlated with inequality.

The final control variables are loosely associated with the effects of ‘globalization’ on income inequality. Southern imports is the natural log of the total sum of imports from non-OECD countries reported as a percent of national GDP.¹² This variable is from the United Nations COMTRADE database and is expected to return a positive association with the dependent variable. FDI outflow is the natural log of the total outward-bound foreign direct investment calculated as a percent of GDP. This covariate is from the United Nations Conference on Trade and Development (UNCTAD) available online at http://unctad.org/en/pages/statistics.aspx and should augment inequality. Finally, inward migration is roughly approximated by calculating the difference between total population increase, on the one hand, and the natural rate of population growth, on the other (Alderson and Nielsen, 2002). This variable is expressed as a ratio of 1000 and should increase inequality.¹³

Methods

The information compiled for this investigation takes the form of a time-series cross-sectional (TSCS) dataset across 25 rich OECD countries for the years 1980 to 2008.¹⁴ This yields the current study 501 total country-year observations.¹⁵ It is well documented that ordinary least squares (OLS) regression is inappropriate for TSCS data. A major issue of OLS with TSCS data is the problem of heterogeneity bias whereby unobserved time-invariant unit-specific variables may be correlated with the observed covariates resulting in biased estimates of the coefficients. TSCS data are also susceptible to autocorrelation and heteroskedasticity especially when the units are nation-states. In this context, the application of OLS regression produces estimates that are neither efficient nor the best linear unbiased estimator.

To deal with these shortcomings, researchers often employ fixed effects model (FEM) or random effects model (REM). FEM possesses the advantage of controlling for all unobservable time-invariant country-level effects making it an extremely useful approach for eliminating spurious relationships. Most crucial is that FEM allows researchers to examine change over time within-countries, rather than between-countries as is the case with REM, which is a major focus of the current study. However, a major limitation of FEM is its inability to analyze time-invariant or slowly changing covariates. Important is that slowly changing variables often lose their statistical significance in the context of FEM. This is noteworthy since this investigation focuses on measures that historically change very slowly over time such as the demographic patterns of employment. In contrast, REM is a better equipped and more efficient estimator of slowly changing covariates. Unfortunately, REM is incapable of controlling for within-country unobservables and produces unreliable estimates when its error terms are correlated with the variables included in the regression models (Halaby, 2004; Wooldridge, 2002).

Given these considerations, a number of preliminary steps are taken to estimate the most optimal models possible. As an initial step, Hausman tests are performed to assess whether heterogeneity bias is a concern in the current investigation. These diagnostics reveal that there is a significant difference in the coefficients produced by FEM and REM which, by extension, shows that the former is the preferred technique of choice. Wooldridge tests reveal that all reported models contain a significant amount of autocorrelation. Preliminary diagnostics also point to a high degree of heteroskedasticity in all reported models indicating the need to control for such processes. In light of these tests, the current study uses Prais-Winsten regression with first-order autocorrelation adjustment, controls for panel-corrected standard errors, and unit-specific fixed effects (Beck and Katz, 1995). The Prais-Winsten models in this study take the following form

Y_{i t} = β_{0} + α_{i} + β_{1} X_{i t - 1} + β_{2} Z_{i t - 1} + ε_{i t}

(1)

where i denotes the country, t the observed year, Y is the dependent variable, α signifies a matrix of country-specific intercepts, X is the ETI, Z is a matrix of control variables, β is the regression coefficient, and ϵ is the error term.¹⁶ In addition, all independent variables are measured at t − 1 to control for potential temporal reverse causality.

In addition, collinearity may be a concern in the pending regressions as some models include sectoral employment statistics together with the ETI. Failure to account for the patterned variation between these independent variables may result in biased estimates. To placate such concerns, robustness checks are performed using two-stage least-squares instrumental variable regression models. This approach is useful in that it controls for potential endogeneity between the right-hand-side variables and also allows for the inclusion of unit-specific fixed effects. The two-stage least-squares regression takes the form

Y_{i t} = β_{0} + α_{i} + β_{1} X_{i t - 1} + β_{2} Z_{i t - 1} + ε_{i t}

(1)

X_{i t - 1} = γ_{0} + α_{i} + γ_{1} W_{i t - 2} + γ_{2} Z_{i t - 2} + υ_{i t}

(2)

where i represents the country, t the observed year, Y the dependent variable, α a matrix of country-specific intercepts, X the ETI, W a vector of instrumental variables, Z a matrix of control variables, β and γ the coefficients, and ϵ and υ the error terms. The ETI is instrumented using various combinations of agricultural employment, industrial employment, service employment, knowledge employment, and female labor-force participation. In addition, all independent variables are measured at t − 1 to control for potential temporal reverse causality. The correlation matrix and summary statistics are available in Table 2.¹⁷

Table 2.

Correlation matrix and summary statistics.

	1	2	3	4	5	6	7	8	9	10	11	12	13	14
1. Income inequality
2. Employment transition	−0.385
3. Knowledge employment	.503	−.161
4. Agricultural employment	−.225	−.246	−.621
5. Sector-dualism	−.315	−.011	−.455	.614
6. Population growth	−.252	−.288	−.330	.314	.229
7. Unemployment rate	.269	−.032	.010	.060	.024	−.227
8. Tertiary enrollment	.292	−.127	.489	−.303	−.238	.018	−.047
9. Female labor-force	.265	−.265	.565	−.325	−.288	−.126	−.427	.336
10. Union density	.123	−.364	.258	−.025	−.217	−.079	−.159	−.113	.559
11. Credit to private sector	.083	.009	.144	−.310	−.175	.132	−.398	.400	.200	−.144
12. Southern imports	.235	−.117	.689	−.377	−.159	−.136	−.064	.404	.321	.049	−076
13. FDI outflow	.136	.095	.436	−.330	−.138	−.243	−.040	.218	.159	.024	.074	.443
14. Inward migration	.181	−.165	.295	−.344	−.334	.148	−.131	.342	.164	−.019	.381	−.101	.225
Mean	39.82	17.36	31.19	1.58	0.53	3.56	1.85	51.91	64.34	38.66	4.38	1.66	2.01	2.31
SD	4.59	6.09	9.48	0.77	1.33	3.57	0.56	18.96	10.78	23.47	0.53	0.53	0.38	3.60
Min	27.65	5.48	5.34	−0.41	−8.41	−5.20	0.47	12.83	33.25	7.05	2.78	−0.29	−0.26	−12.75
Max	49.22	30.22	48.6	3.52	2.87	16.60	3.17	103.5	86.70	99.06	5.76	3.12	3.62	16.90

FDI: foreign direct investment; SD: standard deviation.

Results and discussion

The summary of the Prais-Winsten regressions begins on Table 3. Models 1 to 3 include the ETI and internal development model together with various combinations of control variables. In these specifications, employment in agriculture and population growth return negative and significant association with income inequality, albeit inconsistently. More interestingly, sector-dualism is not significant and is consistently signed in a manner that is counter to expectations. In contrast, the ETI returns a negative and robust correlation with the dependent variable in two of three equations. Although the ETI is signed in the proper direction, it is unable to surpass the minimum significance threshold in Model 3. Models 4 to 6 retest the previous specifications net of knowledge employment. In these regressions, the ETI remains a negative and significant predictor of income inequality. Moreover, knowledge employment returns a positive and robust relationship with income inequality while the internal development model fails to produce any significant results.¹⁸

Table 3.

Fixed effects Prais-Winsten regressions with panel-corrected standard errors.

	1	2	3	4	5	6
Employment transition	−0.096** (−3.17)	−0.089** (−2.99)	−0.046 (−1.65)	−0.096** (−3.21)	−0.090** (−3.03)	−0.055⁺ (−1.88)
Knowledge employment				0.150** (4.89)	0.124** (3.84)	0.089** (2.60)
Agricultural employment	−1.203** (−4.32)	−0.695* (−2.29)	−0.331 (−1.11)	−0.094 (−0.29)	0.066 (0.19)	0.066 (0.20)
Sector-dualism	−0.007 (−0.14)	−0.001 (−0.03)	−0.029 (−0.59)	0.016 (0.34)	0.015 (0.33)	−0.018 (−0.37)
Population growth	0.064 (0.18)	−0.159⁺ (−1.83)	0.239 (0.68)	0.082 (0.23)	−1.218 (−1.43)	0.248 (0.71)
Southern imports		1.426** (3.86)			0.949* (2.48)
FDI outflow		0.064 (0.42)			0.022 (0.15)
Inward migration		0.174⁺ (1.78)			0.137 (1.44)
Unemployment rate			1.951** (5.91)			1.886** (5.71)
Tertiary enrollment			0.028* (2.43)			0.018 (1.54)
Female labor-force			0.131** (3.25)			0.110** (2.70)
Union density			−0.004 (−0.17)			0.004 (0.17)
Credit to private sector			0.665⁺ (1.76)			0.616 (1.63)
Constant	43.420** (33.57)	41.106** (27.06)	24.977** (7.39)	36.709** (20.93)	36.429** (20.00)	23.359** (6.98)
Wald chi²	1943.68	2002.84	1367.90	1448.48	1583.17	1499.41
R ²	.912	.916	.924	.917	.919	.925
Observations	476	476	476	476	476	476
Countries	25	25	25	25	25	25

FDI: foreign direct investment.

Dependent variable is the Gini coefficient; t-values are in parentheses.

p < .10, *p < .05, **p < .01.

There are three observations of note thus far. First, there is no evidence to support the idea that sector-dualism, arguably the most popular measure of the Kuznetsian framework, is a significant determinant of income inequality in postindustrial societies. In fact, most models return a negative correlation for sector-dualism which directly contradicts the expectations of the literature. Second, the ETI returns a negative and significant link with income inequality in a majority of models. These results demonstrate the usefulness of this variable in assessing distributional outcomes, especially when compared to the more classical measurements of the Kuznetsian framework. Finally, knowledge employment returns a positive and significant association with income inequality net of the ETI. Since the ETI controls for the distributional impact of between-sector employment trends, these findings suggest that the knowledge sector produces a significant within-sector effect on income inequality that is unique from its between-sector effect.

Table 4 reanalyzes different combinations of the previously regressed covariates in order to further scrutinize the link between the ETI and income inequality. Model 7 shows that the ETI is a negative and significant predictor of income inequality net of all control variables excluding female labor-force participation. According to Model 8, the introduction of female labor-force participation substantially reduces ETI’s significant association with the dependent variable, which declines from p < .05 to p < .10. In Model 9, the inclusion of knowledge employment augments the ETI’s significant connection with income inequality. However, the ETI is still only able to surpass the p < .10 significance threshold when female labor participation is included in the regression. Five control variables are significant and signed as anticipated across the nine models: southern imports, unemployment rate, and female labor-force participation are positive and significant in every model; inward migration is positive and significant in one of five models; and credit to the private sector is positive and significant in four of five models. Particularly noteworthy is that southern import penetration surpasses the minimum significance threshold in all models analyzed thus far. Furthermore, southern imports breaks the p < .01 significance level in four of the five Prais-Winsten models in which it is included.

Table 4.

Fixed effects Prais-Winsten regressions with panel-corrected standard errors.

	7	8	9
Employment transition	−0.059* (−2.11)	−0.048⁺ (−1.72)	−0.056⁺ (−1.93)
Knowledge employment			0.082* (2.38)
Agricultural employment	−0.444 (−1.50)	−0.259 (−0.84)	0.094 (0.28)
Sector-dualism	−0.041 (−0.88)	−0.032 (−0.68)	−0.022 (−0.46)
Population growth	−0.853 (−0.98)	−0.619 (−0.71)	−0.503 (−0.59)
Southern imports	1.304** (3.40)	1.048** (2.66)	0.942** (2.38)
FDI outflow	0.072 (0.52)	0.072 (0.50)	0.061 (0.42)
Inward migration	0.125 (1.31)	0.094 (0.98)	0.083 (0.88)
Unemployment rate	1.852** (5.63)	2.060** (6.20)	1.973** (5.92)
Tertiary enrollment	0.017 (1.45)	0.016 (1.39)	0.008 (0.71)
Female labor-force		0.100* (2.37)	0.082⁺ (1.93)
Union density	−0.006 (−0.27)	0.000 (0.01)	0.006 (0.29)
Credit to private sector	0.877* (2.47)	0.718⁺ (1.92)	0.666⁺ (1.78)
Constant	31.880** (12.71)	25.539** (7.39)	24.188** (7.05)
Wald chi²	1340.28	1456.41	1567.64
R ²	.924	.925	.926
Observations	476	476	476
Countries	25	25	25

FDI: foreign direct investment.

Dependent variable is the Gini coefficient; t-values are in parentheses.

p < .10, *p < .05, **p < .01.

There are at least two noteworthy observations from these final equations. First, these regressions indicate that the effect of between-sector employment transitions on income inequality is partially explained by the recent explosion of female labor-force participation rates. This is shown by the fact that female labor participation drastically diminishes the connection between the ETI and income inequality. Second, these final models also reveal that the ETI’s association with the dependent variable increases when controlling for knowledge employment’s within-sector effect. That is, knowledge employment’s inclusion in the regressions consistently augments the connection between ETI and income inequality across all models.

Table 5 provides robustness checks by retesting the previous equations via two-stage least-squares instrumental variable regressions. This approach helps to placate potential endogeneity concerns surrounding the right-hand side variables. In the presented specifications, the ETI is instrumented using different combinations of agricultural, industrial, service, and knowledge sector employment. In addition, since the Prais-Winsten regressions indicate that between-sector employment shifts may be partially driven by the entry of women into the labor-force, the ETI is also instrumented with female labor participation in Models 12 and 15. In sum, the results remain largely consistent across both statistical techniques as the ETI is a negative and significant predictor of income inequality in five of six reported models. The non-significant finding in Model 11 is due to the downward pressure exerted on the ETI’s significance level by female labor-force participation. Furthermore, knowledge employment returns strong positive associations with income inequality and continues to augment the significance of the ETI. Also noteworthy is the continued underperformance of the internal development model as the crucial sector-dualism variable achieves significance in two of six models but is signed in the opposite direction, i.e., negative.

Table 5.

Fixed effects two-stage least-squares instrumental variable regressions.

	10	11	12	13	14	15
Employment transition	−0.234** (−2.69)	−0.140 (−1.52)	−0.391** (−4.66)	−0.657** (−6.11)	−0.598** (−4.57)	−0.503** (−6.11)
Knowledge employment				0.324** (8.56)	0.375** (6.39)	0.335** (8.16)
Sector-dualism	−0.338* (−2.26)	−0.214 (−1.56)	−0.401** (−2.75)	−0.089 (−0.55)	−0.114 (−0.72)	−0.088 (−0.60)
Population growth	0.113 (0.11)	1.786⁺ (1.83)	2.123⁺ (1.96)	2.304⁺ (1.94)	2.323* (2.06)	2.211* (2.09)
Southern imports	3.210** (7.02)	0.738 (1.42)	1.889** (3.71)	1.120* (2.05)	1.407* (2.31)	1.105* (2.18)
FDI outflow	1.061** (2.73)	0.502 (1.41)	0.872* (2.26)	0.459 (1.08)	0.496 (1.21)	0.424 (1.11)
Inward migration	−0.113 (−0.87)	−0.257* (−2.14)	−0.329* (−2.47)	−0.459** (−3.12)	−0.394** (−2.81)	−0.367** (−2.83)
Unemployment rate		2.809** (5.06)	1.223* (2.50)		0.986 (1.39)	1.487** (3.13)
Tertiary enrollment		0.039** (2.90)	0.057** (4.02)		−0.009 (−0.58)	−0.007 (−0.50)
Female labor-force		0.222** (4.62)			−0.076 (−1.04)
Union density		0.063** (2.66)	0.052* (2.00)		0.082** (2.99)	0.083** (3.18)
Credit to private sector		0.915⁺ (1.95)	1.128* (2.18)		0.700 (1.30)	0.676 (1.33)
Constant	36.994** (23.67)	11.827* (2.48)	29.309** (8.90)	38.202** (22.62)	32.211** (4.99)	26.607** (8.32)
Wald F statistic	20.80	22.01	18.31	18.19	16.89	19.13
Observations	451	451	451	451	451	451
Countries	25	25	25	25	25	25
ETI instrumented with
Agricultural employment	X	X	X	X	X	X
Industrial employment	X	X	X	X	X	X
Service employment	X	X	X	X	X	X
Knowledge employment	X	X	X
Female labor-force			X			X

FDI: foreign direct investment; ETI: employment transition index.

Dependent variable is the Gini coefficient; t-values are in parentheses.

p < .10, *p < .05, **p < .01.

In sum, three observations are of note from the results taken as a whole. First, it is clear that the ETI outperforms the internal development model and sector-dualism. However, it would be inaccurate to interpret this as proof of the irrelevance of the Kuznetsian framework. The ETI simply demonstrates that the agricultural–industrial transition’s impact on income inequality is declining over time, while the service–knowledge transition’s impact is increasing. Second, the results also indicate that knowledge employment is a positive and significant predictor of income inequality when controlling for the ETI. This demonstrates that the wide dispersion of wages in the knowledge sector generates a within-sector effect on income inequality that is not fully captured by the calculation of between-sector employment trends.

Finally, the findings show that female labor-force participation rates partially account for the connection between employment transitions and income inequality.¹⁹ A large body of work on the new economy shows that the recent growth of income inequality is partially due to the devaluation of women’s work. That is, service jobs are typically considered women’s occupations and are devalued vis-à-vis occupations traditionally associated with men (Cech et al., 2011; Tomaskovic-Devey, 1993). Thus, with the explosion of female labor during the past few decades, many women are entering into a service sector wherein the work they perform is devalued prima facie (Cohen and Huffman, 2003; Dwyer, 2013; Kilbourne et al., 1994). When considered in this context, it is not surprising that female labor participation rates account for a large proportion of the ETI’s contribution to income inequality.

Conclusion

According to the Kuznets Inverted-U hypothesis, income inequality rises and falls as the labor-force migrates from lower- to higher-paid forms of employment. Specifically, income inequality expands during the early stages of industrialization as a small proportion of the labor-force is employed in manufacturing. But this inequality eventually declines as workers continue to migrate to the higher-paid segments of the economy. However, while these assertions are correct conceptually, problematic is that the Kuznetsian view of inequality is contingent on the presence of an incessantly expanding industrial sector. If an economy evolves beyond industrial production, the Inverted-U hypothesis fails to hold. In light of this shortcoming, the current study attempts to amend the distributional dynamics outlined by Kuznets. To this end, this study’s cycle of inequality hypothesis presents the claim that economic development is characterized by a constant series of employment shifts that generate a recurrent cycle of income inequality. In other words, just as the previous agricultural–industrial transition stimulated a rise and fall of inequality, so too is the current service–knowledge transition generating a similar dynamic.

To test the cycle of inequality hypothesis, this study compiles a TSCS dataset for 25 OECD countries for the years 1980 to 2008. More importantly, this investigation formulates an ETI that is used to assess the impact of between-sector employment shifts on income inequality. This index ranges from 0 to 100 where higher scores denote a larger gap of employment between the lower- and higher-wage sectors of the economy. This formulation is based on Kuznets’s logic that inequality is low when a majority of the labor-force is employed in the lower-wage sector (larger differential of employment between the higher- and lower-paid sectors), increases as workers begin to migrate to the higher-wage sector (smaller differential of employment), and declines as the transition to the higher-wage sector continues (larger differential of employment). According to the findings, the ETI is a negative and robust predictor of income inequality and outperforms classical measures such as sector-dualism. Moreover, knowledge employment is positive and significantly associated with income inequality when controlling for the index. This shows that both between- and within-sector trends are important determinants of income inequality.

There are at least two findings and shortcomings of this study that offer viable avenues for future research. First, while the ETI is relatively successful in its prediction of income inequality, it is far from perfect. There are two important considerations when measuring between-sector employment dynamics: first, the demographic patterns of employment, and second, the wage gap between the various sectors. This investigation focuses mainly on the former as the ETI captures the labor-force composition component of the Kuznetsian argument.²⁰ However, the index does not directly account for the wage differential between agriculture and industry, on the one hand, and labor-intensive service and knowledge, on the other. This is an important shortcoming of this study since average within-sector wages change alongside shifting demands for different skills.²¹ In fact, the ETI is nearly the exact opposite of sector-dualism, with the former concentrating mainly on the demographics of employment and the latter solely on the wage gap between sectors.²² Thus, future studies should formulate measures that capture both the employment- and wage-based aspects of the Kuznets hypothesis when the information becomes available.

Second, noteworthy is that female labor participation produces a strong downward pressure on the significance of the ETI. It is important to note that the mass entry of women into the labor-force occurred concurrently with the transition from a goods- to service-producing economy. At the same time, women are largely employed in the lower-paid services which demonstrates the gendered nature of work in the new economy (Beckman and Phillips, 2005; Cech, 2013; Dwyer, 2013; Fletcher et al., 2007; Kim and Sakamoto, 2008; Mouw and Kalleberg, 2010; Nelson and Lorence, 1988). By extension, it is possible that between-sector inequality in postindustrial economies is largely due to the persistence of a gendered wage-gap between the knowledge and service sector (Lin and Tomaskovic-Devey, 2013; Moller et al., 2009; Weeden, 2002; Western and Rosenfeld, 2011). Interesting for future studies would be to engage in sectoral decompositions of between- and within-sector wage inequality among women and men using the occupational categories offered in this study.

In closing, the current investigation amends the Kuznets Inverted-U hypothesis. In doing so, it shows that the current service–knowledge transition is generating a rise and fall of income inequality in a manner that is similar to the previous agricultural–industrial transition. However, history is open-ended, and there is no guarantee that previous socio-economic patterns will hold in the future. Only the passage of time will tell whether income inequality will decline in rich countries over the coming decades with the maturation of the knowledge society.

Footnotes

Appendix 1 Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

References

Acemoglu

Autor

(2011) Skills, tasks and technologies: Implications for employment and earnings. In: Ashenfelter

Card

(eds) Handbook of Labor Economics. Amsterdam: Elsevier, pp. 1043–1072.

Adelman

Morris

(1978) Growth and impoverishment in the middle of the nineteenth century. World Development 6: 245–273.

Ahluwalia

(1976) Income distribution and development: Some stylized facts. American Economic Review 66: 128–135.

Alderson

(1999) Explaining deindustrialization: Globalization, failure, or success? American Sociological Review 64: 701–721.

Alderson

Nielsen

(2002) Globalization and the great u-turn: Income inequality trends in 16 OECD countries. American Journal of Sociology 107: 1244–1299.

Autor

Dorn

(2010) Inequality and specialization: The growth of low-skill service jobs and the polarization of the U.S. labor market. MIT Working Paper.

Autor

Levy

Murnane

(2003) The skill content of recent technological change: An empirical exploration. Quarterly Journal of Economics 118(4): 1279–1333.

Barro

(2000) Inequality and growth in a panel of countries. Journal of Economic Growth 5: 5–32.

Beck

Katz

(1995) What to do (and not to do) with Time-Series Cross-Section Data. The American Political Science Review 89: 637–647.

10.

Beckfield

(2006) European integration and income inequality. American Sociological Review 71: 964–985.

11.

Beckman

Phillips

(2005) Interorganizational determinants of promotion: Client leadership and the attainment of women attorneys. American Sociological Review 70: 678–701.

12.

Bell

(1973) The Coming of Post-Industrial Society. New York: Basic Books.

13.

Black

Spitz-Oener

(2010) Explaining women’s success: Technological change and the skill content of women’s work. Review of Economics and Statistics 92(1): 187–194.

14.

Bluestone

(1990) The great u-turn revisited: Economic restructuring, jobs, and the redistribution of earnings. In: Kasarda

(ed.) Jobs, Earnings, and Employment Growth Policies in the United States. Boston, MA: Kluwer, pp. 7–43.

15.

Bluestone

Harrison

(1982) The Deindustrialization of America: Plant Closings, Community Abandonment, and the Dismantling of Basic Industries. New York: Basic Books.

16.

Cech

(2013) Ideological wage inequalities? The technical/social dualism and the gender wage gap in engineering. Social Forces 91: 1147–1182.

17.

Cech

Rubineau

Silbey

. (2011) Professional role confidence and gendered persistence in engineering. American Sociological Review 76: 641–666.

18.

Chase-Dunn

(1998) Global Formation: Structures of the World-Economy. Lanham, MD: Rowan & Littlefield Publishers.

19.

Chenery

Syrquin

(1975) Patterns of Development, 1950–1970. London: Oxford University Press.

20.

Chevan

Stokes

(2000) Growth in family income inequality, 1970–1990: Industrial restructuring and demographic change. Demography 37: 365–380.

21.

Cohen

Huffman

(2003) Occupational segregation and the devaluation of women’s work across U.S. labor markets. Social Forces 81: 881–908.

22.

Davis

(2009) Managed by the Markets: How Finance Re-Shaped America. Oxford: Oxford University Press.

23.

Deininger

Squire

(1996) Inequality and Growth: Results from a New Data Set. Mimeo, World Bank.

24.

Drucker

(1969) The Age of Discontinuity. New York: Harper & Row.

25.

Drucker

(1993) Post-Capitalist Society. New York: Harper Business.

26.

Dwyer

(2013) The care economy? Gender, economic restructuring, and job polarization in the U.S. labor market. American Sociological Review 78: 390–416.

27.

Fernandez-Macias

(2012) Job polarization in Europe? Changes in the employment structure and job quality, 1995–2007. Work and Occupations 39: 157–182.

28.

Fletcher

Boden

Kent

. (2007) Performing women: The gendered dimensions of the UK new research economy. Gender, Work, and Organization 14: 433–453.

29.

Fligstein

Shin

(2007) Shareholder value and the transformation of the U.S. economy, 1984–2000. Sociological Forum 22: 399–424.

30.

Florida

(2002) The Rise of the Creative Class: And How It Is Transforming Work, Leisure, Community, and Everyday Life. New York: Basic Books.

31.

Freeman

Soete

(1997) The Economics of Industrial Innovation. New York: Routledge.

32.

Goldstein

(2012) Revenge of the managers: Labor cost-cutting and the paradoxical resurgence of managerialism in the shareholder value era, 1984 to 2001. American Sociological Association 77: 268–294.

33.

Gordon

(1996) Fat and Mean: The Corporate Squeeze of Working Americans and the Myth of Managerial Downsizing. New York: Free Press.

34.

Halaby

(2004) Panel models in sociological research: Theory into practice. Annual Review of Sociology 30: 507–544.

35.

Harrison

Bluestone

(1988) The Great U-Turn: Corporate Restructuring and the Polarizing of America. New York: Basic Books.

36.

Howkins

(2001) The Creative Economy: How People Make Money from Ideas. New York: Penguin Press.

37.

Huber

Nielsen

Pribble

. (2006) Politics and inequality in Latin America and the Caribbean. American Sociological Review 71: 943–963.

38.

Jerzmanowski

Nabar

(2013) Financial development and wage inequality: Theory and evidence. Economic Inquiry 51: 211–234.

39.

Kalleberg

(2011) Good Jobs, Bad Jobs: The Rise of Polarized and Precarious Employment Systems in the United States, 1970–2000. New York: Russell Sage Foundation.

40.

Katz

(2000) Technological change, computerization, and the wage structure. In: Brynjolfsson

Kahin

(eds) Understanding the Digital Economy: Data, Tools, and Research. Cambridge: MIT Press, pp. 217–244.

41.

Kilbourne

Farkas

Berton

. (1994) Returns to skill, compensating differentials, and gender bias: Effects of occupational characteristics on the wages of white women and men. American Journal of Sociology 100: 689–719.

42.

Kim

Sakamoto

(2008) The rise of intra-occupational wage inequality in the United States, 1983 to 2002. American Sociological Review 78: 129–157.

43.

Korzeniewicz

Moran

(2005) Theorizing the relationship between inequality and economic growth. Theory and Society 34: 277–316.

44.

Korzeniewicz

Moran

(2009) Unveiling Inequality: A World-Historical Perspective. New York: Russell Sage Foundation.

45.

Kristal

(2013) The capitalist machine: Computerization, workers’ power, and the decline of labor’s share in U.S. industries. American Sociological Review 78: 361–389.

46.

Krueger

(1993) How computers have changed the wage structure: Evidence from microdata, 1984–1989. Quarterly Journal of Economics 108: 33–60.

47.

Kuznets

(1955) Economic growth and income inequality. The American Economic Review 45(1): 1–28.

48.

Lecaillon

Paukert

Morrisson

. (1984) Income Distribution and Economic Development: An Analytical Survey. Geneva: International Labour Office.

49.

Lee

(2005) Income inequality, democracy, and public sector size. American Sociological Review 70: 158–181.

50.

Lee

Kim

Shim

(2011) The limit of equality projects: Public-sector expansion, sectoral conflicts, and income inequality in postindustrial economies. American Sociological Review 78(1): 100–124.

51.

Lee

Nielsen

Alderson

(2007) Income inequality, global economy and the state. Social Forces 86(1): 77–111.

52.

Levy

Michel

(1991) The Economic Future of American Families: Income and Wealth Trends. Washington, DC: Urban Institute Press.

53.

Levy

Murnane

(1992) U.S. earnings levels and earnings inequality: A review of recent trends and proposed explanations. Journal of Economics Literature 30: 1333–1381.

54.

Lin

Tomaskovic-Devey

(2013) Financialization and U.S. income inequality, 1970–2008. American Journal of Sociology 118: 1284–1329.

55.

Lorence

Nelson

(1993) Industrial restructuring and metropolitan earnings inequality, 1970–1980. Research in Social Stratification and Mobility 12: 145–185.

56.

Machlup

(1962) The Production and Distribution of Knowledge in the United States. Princeton, NJ: Princeton University Press.

57.

Mahutga

Kwon

Granger

(2011) Within-country inequality and the modern world system: A theoretical reprise and empirical first step. Journal of World-Systems Research 17(2): 279–307.

58.

Mann

Riley

(2007) Explaining macro-regional trends in global income inequalities, 1950–2000. Socio-Economic Review 5: 81–115.

59.

Misra

Boswell

(1997) Dutch hegemony during the age of mercantilism. Acta Politica: International Journal of Political Science 32(2): 174–209.

60.

Modelski

Thompson

(1996) Leading Sectors and World Powers: The Coevolution of Global Politics and Economics. Colombia, SC: University of South Carolina Press.

61.

Moller

Alderson

Nielsen

(2009) Changing patterns of income inequality in U.S. counties, 1970–2000. American Journal of Sociology 114(4): 1037–1101.

62.

Moran

(2005) Kuznets’s inverted U-curve hypothesis: The rise, demise, and continued relevance of a socioeconomic Law. Sociological Forum 20(2): 209–244.

63.

Mouw

Kalleberg

(2010) Occupations and the structure of wage inequality in the United States, 1980s to 2000s. American Sociological Review 75: 402–431.

64.

Nelson

Lorence

(1988) Metropolitan earnings inequality and service sector employment. Social Forces 67: 492–511.

65.

Nielsen

(1994) Income inequality and industrial development: Dualism revisited. American Sociological Review 59: 654–677.

66.

Nielsen

Alderson

(1997) The Kuznets curve and the great u-turn: Income inequality in U.S. counties, 1970 to 1990. American Sociological Review 62(1): 12–33.

67.

Perez

(2003) Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages. Northampton, MA: Edward Elgar Publishing.

68.

Piketty

(2014) Capital in the Twenty-First Century. Cambridge, MA: Harvard University Press.

69.

Rohrbach

(2009) Sector bias and sector dualism: The knowledge society and inequality. International Journal of Comparative Sociology 50: 510–536.

70.

Sanderson

(1999) Social Transformations. Lanham, MD: Rowman & Littlefield Publishers.

71.

Sassen

(1996) Service employment regimes and the new inequality. In: Mingione

(ed.) Urban Poverty and the Underclass: A Reader. Oxford: Blackwell Publishers, pp. 64–82.

72.

Sassen

(1998) Globalization and Its Discontents. New York: New Press.

73.

Schumpeter

(1939) Business Cycles. New York: McGraw-Hill.

74.

Solt

(2009) Standardizing the world income inequality database. Social Science Quarterly 90(2): 231–242.

75.

Stanback

(1979) Understanding the Service Economy. Baltimore, MD: Johns Hopkins University Press.

76.

Tomaskovic-Devey

(1993) The gender and race composition of jobs and male/female, white/black pay gaps. Social Forces 72: 45–76.

77.

Tsai

(1995) Foreign direct investment and income inequality: Further evidence. World Development 23(3): 469–483.

78.

US Department of Labor (2012) Bureau of Labor Statistics. Washington, DC: US Department of Labor.

79.

Volscho

Kelly

(2012) The rise of the super-rich: Power resources, taxes, financial markets, and the dynamics of the top 1 percent, 1949 to 2008. American Sociological Review 77(5): 679–699.

80.

Wallerstein

(1984) The Politics of the World Economy. Cambridge: Cambridge University Press.

81.

Weeden

(2002) Why do some occupations pay more than others? Social closure and earnings inequality in the United States. American Journal of Sociology 108: 55–101.

82.

Western

Rosenfeld

(2011) Unions, norms, and the rise in U.S. wage inequality. American Sociological Review 76(4): 513–537.

83.

Williamson

Lindert

(1980) American Inequality: A Macroeconomic History. New York: Academic Press.

84.

Wooldridge

(2002) Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: MIT Press.

85.

World Bank (2013) World Development Indicators (online). Washington, DC: World Bank.

86.

Wright

(1997) Class Counts: Comparative Studies in Class Analysis. Cambridge: Cambridge University Press.