Comparing past and present wage inequality in two globalisation periods

Abstract

This paper compares past and present globalisation with an aim to highlighting the different factors that drove wage inequality then and those which are doing so now. We have constructed a ratio of wage inequality for 15 countries in the first period of globalisation (1870–1913) and the subsequent period of deglobalisation (1914–1930) and then compare this pattern to wage inequality in the 1980s and 1990s. We propose that the difference in wage inequality trends for the two globalisation periods is due to migration and institutional factors (education and labour market institutions). These factors offset the increase in wage inequality produced by globalisation and technological change in the past, but do not appear to be acting in the present.

Keywords:

• globalisation
• wage inequality
• education
• labour market institutions

JEL Classification:

• N30
• J31
• F6
• I24

Acknowledgements

We are grateful to the Spanish Ministry of Education (SEJ2004-07402 and ECO2012-33337) and the University of Valencia (UV-AE-09-5846) for their financial support and Michael Huberman and the CIREQ, Université Montréal for the research stay of Concha Betrán during the elaboration of the paper. This paper has been presented at CEA (Ottawa, June 2011). We would like to thank Doug MacCallas, the anonymous referees and the editors for their comments.

Notes

¹Atkinson, ‘Top Incomes’ (2002).

²O'Rourke/Williamson, Globalisation (1999).

³O'Rourke/Williamson, Globalisation (1999) and O'Rourke et al., ‘Factor Price’ (1996).

⁴Lindert/Williamson, ‘Reinterpreting Britain's Social Tables’ (1983).

⁵Lindert, ‘Unequal English Wealth’ (1986).

⁶Williamson, ‘Globalisation’ (1996).

⁷O'Rourke/Williamson, Globalisation (1999).

⁸Anderson, ‘Globalisation’ (1999).

⁹Women have been excluded because of a lack of data, but most of them were unskilled workers employed in agriculture, the textile and apparel sectors or domestic service. Hence, the inclusion of women would have increased wage inequality.

¹⁰We are also conditioned by data sources. It is very difficult to analyse wage inequality for this period other than in terms of differences in wages between the various labour market groups and patterns of wage inequality are based on occupational averages. See Margo (1999), ‘History of Wage Inequality’.

¹¹Williamson, ‘Globalisation’ (1996).

¹²Our data indicate that building wages are not always a good indicator of the evolution of wages for industry as a whole. A more detailed discussion about whether the building sector can be used to represent the evolution of wages in industry as a whole will be provided upon request.

¹³Lemieux, ‘Changing Nature’ (2008).

¹⁴Betrán et al., ‘Explaining UK Wage Inequality’ (2010).

¹⁵Most of the Old World countries were sending countries along the nineteenth century, with some exceptions such as France. The timing of emigration was different among countries, for example, Germany and the UK had the higher emigration rates in the first part of the nineteenth century whereas Italy and Spain in the late nineteenth century.

¹⁶In the case of Australia and Germany, we have data on wage inequality for various sectors, but not for the whole period and for this reason we have used wage inequality in the building industry to interpolate for the years we lack these data (see Appendix 1).

¹⁷‘The Historical Statistics of Canada’ provides other data for the building industry and also for the ratio ‘supervisory and office workers/production manufacturing’. However, these data have been criticised by many researchers who consider that the Mackinnon series (see Appendix 1) on the railway industry is more representative of the evolution of wages in the whole economy. Green and Green Canada's Wage Distribution (2007) offer an estimation of wage inequality for 1911, 1921 and 1931. Unfortunately, they do not have data for the globalisation period. For the years, they do have data, they obtain a similar pattern to the Mackinnon series: a decrease in wage inequality from 1911 to 1921 and a recovery from 1921 to 1931.

¹⁸The only case where we have not obtained a clear increase in wage inequality in this period is Brazil.

¹⁹Japan also experienced a very smooth rise in wage inequality, but was the country with the lowest level of wage inequality. Moreover, most of the increase occurred at the beginning of the period.

²⁰For the countries for which we have data on wage inequality, we obtain the same pattern during the period 1930–1939 as before. This is the case of Australia, Denmark, Germany, Japan and the USA. The data for the UK are not as complete because we only have data for 1933 and 1935 and observe that wage inequality remains almost constant.

²¹O'Rourke/Williamson, Globalisation (1999).

²²Anderson, ‘Globalisation’ (2001).

²³Earnings include overtime pay, bonuses and gratuities.

²⁴OECD labour market statistics database, 2003. The data are mainly from household and enterprise surveys. The earnings series corresponds to different time measures: hourly, daily, weekly, monthly, and annual earnings, which can affect the comparisons of levels in different countries. We have not extended the period to more recent years because the OECD has changed the database resulting in only a few countries having homogenous data from 1980 to the present. Furthermore, the other sources used for the non-OECD countries are only for the 1980s and 1990s.

²⁵We have to take into account that part-time jobs are mainly unskilled. With respect to woman, there has been a great increase in women's labour market participation in low-skilled sectors and very often in part-time employment. See, Gregory/Connolly, ‘Part-Time Work’ (2008).

²⁶Lemieux, ‘Changing Nature’ (2008), 21–48.

²⁷The problem with this ratio is that very different occupation categories are included in non-production workers and the composition of these categories has changed over time.

²⁸The World Bank provides data on income inequality. However, we are interested in analysing inequality in the labour market and income inequality includes other sources of inequality, such as land rents, etc., which are not considered in this paper.

²⁹Lemieux, ‘Changing Nature’ (2008), 28.

³⁰We consider as developing countries those with an average per capita income below $20.000 at the end of the twentieth century.

³¹This is a significant limitation because we have not been able to consider African countries, among others.

³²For Spain, we only have data for 1995. There is another source: social security records, from Bover et al., ‘Distribution of Earnings’ (2002), 3–53, and we observe an increase in wage inequality from 1980 to 1987.

³³Wood, North-South Trade Employment (1994).

³⁴Knight/Song, ‘Urban Wage Inequality’ (2003).

³⁵Portela, ‘Wage Inequality’ (2002).

³⁶See Robertson, ‘Regional Integration’ (2002), 260–291 and Robertson, ‘Relative Prices’ (2004).

³⁷Pavcnik, ‘What Explains Skill Upgrading’ (2003).

³⁸Gonzalez/Miles, ‘Wage Inequality’ (2001).

³⁹Goldberg/Pavnick, ‘Distributional Effects of Globalisation’ (2007).

⁴⁰See O'Rourke/Williamson, Globalisation (1999) and Anderson, ‘Globalisation’ (2001).

⁴¹Betrán/Pons, ‘Skilled and Unskilled Labour Wage Differentials’ (2004).

⁴²Betrán et al., ‘Explaining UK Wage Inequality’ (2010).

⁴³See Goldin/Katz, ‘Technology’ (1996), Goldin/Katz, ‘Origins of Technology-skill Complementarity’ (1998), Goldin/Katz, ‘US Education Leadership’ (2001) and Goldin/Katz, Race between Education and Technology (2008).

⁴⁴See James/Thomas, ‘Industrialization’ (2000), or Harley, ‘Skilled Labour’ (1974), for the past.

⁴⁵Landes, Unbound Prometheus (1969).

⁴⁶Machin/van Reen, ‘Technology’ (1998), Aghion/Howitt, ‘Wage Inequality’ (2002), Aghion et al., ‘General Purpose Technology’ (2002), Card/DiNardo, ‘Skill-Biased Technological Change’ (2002). Autor et al., ‘Trends in US Wage Inequality’, (2008) and Acemouglu, ‘Technical Change’ (2002).

⁴⁷See Wood, North-South Trade Employment (1994), Slaughter, ‘International Trade’ (1998), Feenstra, Impact of International Trade (2000) and Haskel/Slaughter, ‘Trade’ (2001).

⁴⁸See Wood, ‘Globalisation’ (1998) and Goldberg/Pavcnik, ‘Distributional Effects of Globalisation’ (2007). This last paper insists on the fact that the increased wage inequality in many developing countries that are relatively abundant in unskilled labour seems inconsistent with the Heckscher–Ohlin theory.

⁴⁹Feentra/Hanson, The impact of Outsourcing (1999).

⁵⁰See Hatton/Williamson, Age of Mass Migration (1998).

⁵¹Timmer/Williamson, ‘Immigration Policy’ (1998).

⁵²Levy/Temin, ‘Inequality’ (2007).

⁵³Huberman/Meissner, ‘Riding the Wave of Trade’ (2010).

⁵⁴Wright, ‘Productivity Growth’ (2006), 139–160.

⁵⁵Hobsbawm, ‘New Unionism’ (1985), 23.

⁵⁶Huberman/Lewchuk, ‘European Economic Integration’ (2003).

⁵⁷Huberman/Lewchuk, ‘European Economic Integration’ (2003).

⁵⁸Easterlin, ‘Why Isn't the Whole World Developed?’ (1981).

⁵⁹Lindert, Growing Public (2004).

⁶⁰Goldin/Katz, ‘US Educational Leadership’ (2001) and Goldin/Katz, Race between Education and Technology (2008).

⁶¹Williamson, ‘Globalisation’ (1996).

⁶²The unobserved effect is eliminated when the mean values of the variables in the observations on a given country are calculated and subtracted from the data for that country. We focus on within-country variation over time or the relationship within each of our countries between the endogenous variable and the explanatory variables. The model is called a within-group regression for that reason, because it is explaining the variations in the means of the explanatory variables for the group of observations relating to a given country.

⁶³In this case, the unobserved effects or variables are randomly distributed and are treated as random variables. In addition, these unobserved effects are distributed independently of all the explanatory variables. The random effects model is estimated by Generalised Least Squares.

⁶⁴This trend is also included in the estimations for the present such as Autor et al., ‘Trends in US Wage Inequality’ (2008).

⁶⁵We have tested for the best estimation between random effects and OLS estimations. The Breusch-Pagan Lagrange Multiplier (LM) test is designed to test random effects. The null hypothesis of the one-way random group effect model is that variances of groups are zero. If the null hypothesis is not rejected, the pooled regression model is appropriate. If not, the random effect model would be. As we have rejected the null hypothesis, we use the random effects model.

⁶⁶Autor et al., ‘Trends in US Wage Inequality’ (2008).

⁶⁷Goldin/Katz, ‘US Educational Leadership’ (2001) or Boudarbat et al., ‘Returns to Human Capital’ (2010).

⁶⁸Lemieux, ‘Post-secondary Education’ (2006) or Lindley/Machin, ‘Rising Wage Inequality’ (2011).

⁶⁹Wälde, ‘Egalitarian and Elitist Education’ (2000).

⁷⁰Berges, ‘Inequalities in Human Capital’ (2009).

⁷¹Card et al., ‘Unions’ (2004). In the same line, Gosling and Lemieux maintain that the reason behind wage inequality growth in the UK outpacing that in the USA was the much steeper decline in unionisation in this country. See Golsing/Lemieux, ‘Labor Market’ (2007), 275–312. Also, Machin states that the countries with the largest increase in wage inequality are those where the influence of labour market institutions on wage setting has declined the most. See Machin, ‘Wage Inequality’ (1996).

⁷²Lee, ‘Wage Inequality’ (1999) and Lemieux, ‘The Changing Nature’ (2008).

⁷³Autor et al., ‘Trends in US Wage Inequality’ (2008).

⁷⁴Gonzalez/Miles, ‘Wage Inequality’ (2001).