A New Kuznetsian Dynamic: The Knowledge Economy and Income Inequality in the United States, 1917–2008

Abstract

The rise of the knowledge economy resulted in higher levels of income inequality in the United States and forced many to question the Kuznets Inverted-U hypothesis. However, this study argues that the establishment of a knowledge economy does not negate the importance of employment shifts for income inequality. Instead, the expansion of knowledge employment alters the major sectors that are responsible for the overall distribution of income. To this end, this article presents the key argument that the current service–knowledge transition impacts income inequality trends, of today, in a way that is similar to the agricultural–industrial transition, of the past. According to the autoregressive conditional heteroskedasticity regressions, the agricultural–industrial transition returns stronger associations with income inequality in the United States before 1950. The agricultural–industrial transition's impact diminishes thereafter as the service–knowledge transition shares a more robust association with income inequality after 1980.

Keywords::

• income inequality
• labor-force trends
• knowledge economy
• Kuznets Inverted-U
• Great U-Turn

ACKNOWLEDGMENTS

The author thanks Rob Clark and Anthony Roberts for their helpful comments on previous iterations of this article. This article was presented at the 2013 annual meetings of the sociology of development conference in Salt Lake City, Utah.

NOTES

Notes

¹ Top income shares do not directly measure the overall distribution of earnings. But it allows for a more temporally expansive view of inequality as the measurements required to construct a U.S. Gini coefficient are unavailable before the 1940s. Additionally, unreported comparisons of the 90-10 ratio to the Gini coefficient return a correlation of 0.96.

² See Kalleberg (2011) for a description of this process.

³ The 1950s period is important as employment in agriculture and industry, as a percent of total employment, fell below 50 percent.

⁴ Others use the term “postindustrial” or “post-Fordist” economy (Hirschorn 1984; Block 1990).

⁵ There is an interesting, if not complementary, line of research which shows that knowledge workers are attracted/fostered by the availability of public amenities (e.g., Glaeser 1998; Clark 2003; Shapiro 2006).

⁶ Interestingly, Florida's (2002) estimate of knowledge employment is strikingly similar to that of the expert-managerial class estimated by Wright (1997) and Barley (1996).

⁷ There is a large literature on the role of skill-biased technological change as the primary cause of rising inequality (Dustmann et al. 2009; Autor and Dorn 2010; Black and Spitz-Oener 2010; Michaels et al. 2010). The evidence includes the similar timing of increasing wage inequality and the invention of personal computers (Katz 2000), while others observe that educated and better paid workers are more likely to utilize computers in the workplace (Krueger 1993). But others note that inequality rises years before the widespread availability of computers (e.g., Card and DiNardo 2002).

⁸ This conversion is done by logging all variables and using the difference function in ARIMA.

⁹ The dependent variable in this study includes returns to capital investments. However, noncapital gains included estimates of the dependent variable were also tested but this did not substantially alter the findings. This analysis is available on request.

¹⁰ Many use southern imports as opposed to total imports. But an unreported analysis shows that this variable is not a significant predictor of United States income inequality.

¹¹ Other time periods were tested but the current cutoffs return the most robust associations.

¹² Preliminary diagnostics and correlation matrix is available on request.

¹³ Some scholars claim that inequality is less systematically determined today that in the past (e.g., Burtless and Jencks 2003). This argument implies that model residuals may grow more dispersed over time resulting in time-contingent heteroskedastic errors. However, an unreported supplementary analysis, which is available on request, reveals that the residuals of the models specified in this investigation does not produce consistently more dispersed residuals over time.

¹⁴ There should be less concern that correlated independents are biasing the results since the indicators are log-differenced. Nevertheless, the exclusion of unionization substantially enhances the significance of international imports and, to a lesser extent, female labor participation.

¹⁵ For an earlier review of the literature against Kuznets, see Fields and Jakubson (1994).

¹⁶ This is important as Kuznets offers a vague explanation regarding the point at which inequality will start to decline during the transition from “traditional” to “modern” forms of employment.