Income inequality has increased worldwide in recent years. This column discusses the role of technological progress, globalization and the liberalization of labour-market institutions in this growing inequality. The liberalization of labour market institutions has made labour markets more flexible and created many jobs. But beyond a certain point, the net effect of further liberalization might be negative for society.

Over the past couple of years, the OECD has drawn attention to the rapidly widening income dispersion in OECD countries. The recent publication of Thomas Piketty’s “Capital in the 21st Century”, gave new impetus to this debate.

There were large shifts in the distribution of total income, both from labour to capital, and within the labour share – from low- to high-earning workers. This growing inequality has a large impact on many aspects of society, from diverging educational opportunities, to the distribution of health and overall well-being. Though there are marked differences between countries, the rise in inequality has been a general phenomenon. Hence, there is a quest for more inclusive strategies that allow larger parts of society to benefit from the growth of GDP.

Causes of income inequality

There has been much debate on the causes of this trend. At the turn of the century, skill-biased technological progress was considered to be the main culprit. The skill bias in the change in production technology reduced the demand for high school dropouts and raised that for college graduates. To the extent that the education system could not keep pace with the increased demand for graduates, the skill premium soared.

The rapid pace of technological progress points towards an increasingly winner-takes-all society, where early entrants in new industries capture an incredible amount of rents, with Microsoft, Google and Facebook as the most extreme examples. This contributed to a steady shift in the distribution of income from labour to profits.

More recently, the attention shifted to the role of globalization. Where the share of the BRIC countries in the world economy was too small to account for the large shifts in relative wages during the nineties, this share has grown so rapidly in the last two decades that its impact on the income distribution can no longer be ignored. However, this impact of globalization is more subtle than an outright increase in inequality. For the BRIC countries, globalization is equivalent to a large increase in GDP and, hence, a large improvement in the standards of living for the majority of their population.

For the developed countries, the effect of globalization on the income distribution is more subtle. The wage distribution becomes increasingly skewed to the right. Low and medium educated workers earn more and more similar wages, forming a large mass on the left side of the income distribution, while the share of the top 10% in total income skyrocketed, leading to a long and fat right left tail of the distribution.

Specifically, the middle class loses (see Autor et al. 2013). The middle class is employed exactly in the type of jobs that both face fierce global competition and are highly vulnerable to automation. Demand for workers with intermediate levels of education, therefore, falls and hence their wage surplus above workers with lower education. Despite the overall growth, US real wage of the median worker has fallen since 1990.

The upper class commands specialized human capital that is in high demand in a technology-driven world, whereas the lower class is mainly employed in personal services that are non-tradable and cannot be imported from the BRIC countries. Moreover, jobs in these industries are less sensitive to technological innovation than jobs in manufacturing.

From a policy perspective, this diagnosis poses new challenges. The old story for emancipation relied on investment in human capital; for the lower strata of society, education was the best means for improving their position. The future of your children was safeguarded most effectively by better education – make sure that they spend all effort to obtain the highest possible degree. That story no longer applies for everybody. It still works for the upper tale of the distribution of educational attainment, since the return to obtaining a degree from a good university has gone up. It no longer works for the lower tale of the distribution, since the return to the completion of high school, or adding some years of college, has gone down.

However, trade and technology are not the only explanations for the dramatic increase in inequality over the past couple of decades.

The liberalization of labour-market institutions since 1980 has also contributed substantially. DiNardo, Fortin and Lemieux (1996) have shown that the fall in minimum wages and the demise of the union increased inequality in the United States during the 1980s. Later work by Lee (1999) and Teulings (2003) suggests that a large part – if not all – of the rise in wage inequality in the lower half of the distribution is due to the fall in minimum wages during that period. This applies in particular for females, who earn lower wages anyway, and for whom the minimum wage therefore has a larger impact.

Starting from the low levels of the minimum wage that prevail in the US currently, the job losses of an increase in that minimum are limited. This fits the recent experience in the UK, where the introduction of a minimum wage had a substantial effect on wage inequality, but hardly any effect on the chances of low skilled workers to find a job. It might be different in France, where minimum wages are much higher to begin with. In that case, further increases in the minimum wage have large detrimental effects on the job opportunities for low-skilled workers.

Minimum wages – provided that they are not set too high – therefore have a large effect on wage inequality, and a relatively small effect on employment. The obvious question then is: what wage would be not too high?

I use as a rule of thumb that at most 4% of the workers may actually earn the minimum wage. If more than 4% earns the minimum, job loss becomes substantial. However, there is no empirical evidence to support the rule of thumb other than the differences in outcomes between the US and France. Regrettably, the strategy of a limited increase in the minimum wage reaches this critical point of 4% earlier now that the wage differential between low and medium educated workers has gone down, and more workers earn wages not that far from the minimum wage.

Institutions and the labour market

Why are institutions so important on the labour market? Why would the free market not lead to the best possible outcome?

We gradually start to understand why the institutions for wage setting matter so much. The standard auction model taught in economics 101 is a bad representation of how labour markets actually operate. The auction model assumes that all job seekers and all firms with vacancies meet at the same place and time. A hypothetical auctioneer calls a potential wage rate, records potential supply and demand at that wage, and then adjusts his call until he arrives at an equilibrium rate that sets supply equal to demand.

Reality looks different. Individual job seekers and firms meet and bargain bilaterally on an acceptable wage. If they fail to reach agreement then both continue search for other options. The point is that there is no auctioneer to set the wage. The worker and the firm have to agree among themselves. Hence, it is critical which of the two parties has the best bargaining skills – the worker or the firm? There is no guarantee whatsoever that the free market yields a proper distribution of bargaining power from the point of view of social well-being.

When the theoretical apparatus for analysing these models was first developed by Pissarides (1990) and Burdett and Mortensen (1988) – who received the Nobel prize for this work – it looked as though, by a kind of divine miracle, the actual distribution of bargaining power was reasonably in line with the social optimum. Later on, Stevens (2004) showed that firms can save on their wage bill by using deferred compensation schemes. Workers get paid low wages initially, only to receive higher wages after staying at the same firm for a couple of years. If the worker quits before, the firm never has to pay these higher wages.

Deferred payment binds a worker to the firm and, therefore, reduces the threat of outside competition. No practical man would be surprised by this result because this is what we see employers do every day. Most wage contracts contain extensive experience and tenure scales. However, what is surprising is that these schedules shift the balance of power in favour of the employer beyond what is desirable from societal point of view. Hiring a worker becomes too attractive. Firms start poaching workers from each other, leading to a waste of resources on recruitment activity, excess job mobility, and training cost to make workers familiar with their new job, from which they are likely to quit shortly afterwards by new incoming job offers anyway. This change in the balance of power might also have contributed to the shift of the distribution of total income from labour to capital.

Together with Gautier and Van Vuuren (2010), I showed that even when firms don’t use Stevens’ seniority profiles, but instead offer a fixed flat wage contract, efficiency emerges only when there are strongly increasing returns to scale in a job search, and when firms can commit to pay hiring premiums before even meeting a potential candidate. In all other cases, firms have too much bargaining power.

Robin and Postel-Vinay (2002) have shown that firms can use even more aggressive strategies by paying wage increases only when an outside firm threatens to poach its worker. A firm can then hire apprentices and other young workers for very low starting salaries, waiting for outside offers to these workers before paying any wage increase. Empirical evidence shows that this type of arrangements is used in practice indeed, in particular for low skilled workers. This type of wage contract shifts the balance of power even more in favour of employers, so even further beyond the social optimum than in Stevens’ analysis.

In this type of environment, small institutional details in wage setting can have a large impact on the outcome. Hence, the wave of liberalization of labour-market institutions in the 1980s and 1990s might have a negative impact on society after all. It has made labour markets more flexible and thereby created many jobs, but beyond a certain point, the net effect of further liberalisation might be negative from a societal point of view.

Both the Financial Times and The Economist expressed sympathy for the idea of raising the minimum wage in the United States and introducing it in Germany. This support might be well taken. There is likely to be some kind of an optimum degree of liberalization of labour market institutions. In some instances, the world might have moved beyond that point.

Published in collaboration with VoxEU

Author: Coen Teulings, Professor of Economics, University of Cambridge; and CEPR Research Fellow

Image: A woman walks past graffiti which reads, “1 percent to get rich and 99 percent to get poor”, on the streets of Lisbon August 26, 2012. REUTERS/Hugo Correia