Inequality resources

Charts 1b: New earnings Ginis series and comparisons

Venezuela: Earnings Gini

Comparisons among the inequality series:

For the post 1950 years our series fits reasonably well the trends and fluctuations observed in the household Ginis available, especially in the case of Argentina, Brazil, Chile and Mexico. This gave us some reassurance that the early estimates can also be a good proxy for the “true” income inequality during the first half of the last century. Note that ours reflects inequality before net taxes and transfers, whereas the household Ginis are based on disposable income after net taxes and transfers. This means that our estimates tend to underestimate the observed narrowing in inequality in the last decade on the new century (e.g. in Venezuela).

The comparison between our series and those calculated by FitzGerald show enough differences, both in terms of trends and fluctuation, to justify the construction of the new set of series with superior information ( i.e., substituting real wages for sectoral labour productivity) but with the use of the same methodology. They also do a better job than FitzGerald´s series in matching developments in the household Ginis.

With the exception of Colombia, the new series tend to be consistent with the Williamson ratio (1900-40), particularly by showing an increasing inequality trend to about 1920. This give additional support to claims that the First Globalisation brought about a significant worsening in inequality and that the colonial roots may have played a lesser role than that implied by the institutionalist approach in accounting for high inequality levels observed in the closing decades of the last century.

Regarding the consistency with Bértola et al (2010) estimates for 1920, little can be said because there is no trend to compare with. The authors provide another data point circa 1870, but any interpolation between those two points can be largely misleading. However, there are two potential implications. First, according to our series for Brazil, 1920 appears as an outlier of an otherwise rising trend during the first three decades of the last century or so. This would suggest that 1920 is not a typical year to measure prevailing inequality levels in the country at the time (Bértola et al report a high Gini of 0.597). Secondly, if there is an upward bias in the 1920 estimate, their 1872 estimate for Brazil (0.548) would suggest that the First

11

Globalisation did little to worsen inequality in the country - or even that inequality could have been reduced during the period.4

Trends, peaks and fluctuations

Our estimates indicate a significant inequality reduction during the Great Depression in Colombia and Mexico and, to a lesser extent, in Argentina and Chile. One possible explanation of this result is the combined action of nominal wage rigidity and deflation (all four countries experienced deflation in the early 1930s) boosting real wages at the expense of unemployment. Whereas, this effect tends not to be reflected in the GDP per worker series that in most cases shows a contraction. As a consequence, earnings of group 1 (employers, managers & professionals) – which is calculated as a residual – shrink considerably. But note that because we are using economically active persons rather than employment series to calculate the shares of each occupational group as well as to calculate GDP per worker, our earnings Ginis tend to overestimate a reduction in inequality at times of high unemployment such as during the Great Depression.

The 1970s tend to be dominated by a high level of inequality with a peak in Argentina, Chile and Colombia. The decades of the 1930s, 1940s and the 1970s are sources of fluctuations in inequality. If the narrowing trend in the former can be associated with rigidities and multiple exchange rates (allowing for a currency overvaluation), the worsening trend during the latter period could have been influenced by hyperinflation in Argentina, Brazil and Chile hurting wages. In both cases, it could be argued that the fluctuations responded to unstable macroeconomic policy, and that the effects are not part of a secular trend in inequality.

For the calculation of overall inequality in the region we add the Bértola´s series for Uruguay. This allows us to produce averages - simple and population weighted - for seven Latin American countries (LA-7). We also calculate the average inequality curve for the Southern Cone. This separation is of interest because this group of countries industrialised relatively earlier and could display different inequality-growth dynamics.

4 However, this apparent paradox may not arise with the original 1872 Gini of 0.366 as reported in Bértola et al, 2008, which would indicate a worsening in inequality in 1920 relative to 1872 - even after adjusting for a potential upward bias in their 1920 estimate.

12

The following chart includes three regional averages for the LA-7: simple average and population weighted for the LA-7, and a simple average for the Southern Cone (all series are 5-year moving averages). We also include the summary of our earnings Gini series by country plus Uruguay.

The LA-7 aggregates seem to confirm the presence of an “N” shaped curve for the period 1900-2000 as argued by FitzGerald, with turning points circa 1930 and 1960. So increasing inequality during the earlier decades can equally be attributed to the First Globalisation, the industrialisation process (particularly in the Southern Cone and Mexico) and the dominance of an adverse macroeconomic environment (bust). Then the improvement of inequality that follows could be explained in terms of protectionist policies favouring urban workers, the start of a declining trend of the Kuznets process, or an expansionary macroeconomic environment.

However, the period of narrowing inequality is relatively short-lived, as by the early 1960s the countries tended to return to the early peak in inequality during the 1920s.

Indeed, the explanations behind an “N” shape point to a second turning point around mid 1970, not 1960. Then inequality levels reached a plateau circa the debt crisis that lasted about two decades before beginning to decline with the turn of the new century. When considering the extended period to 2010, the curve turns into an “M” shape. These shapes are more pronounced in the case of the Southern Cone countries, indicating that the overall pattern in the LA-7 is driven by that group of countries.

13

4. Final remarks

We have presented an improved and expanded set of earnings Ginis for the period 19002010. The calculation of this inequality indicator required assembling a consistent set of real wage series for agriculture, manufacturing and the urban sector for our six Latin American countries. This basic data is still preliminary. Though we believe that we are already taking into account most of the data available, there is still room for improvement in some countries and periods (e.g. manufacturing wages in Mexico in the years around 1930, and overall better data for urban wages in the last two decades or so). We may also need to fine tune some of the assumptions made to complete the series when data was lacking (e.g., in Brazil during late 1920s and early 1930s and in Venezuela in early decades). However, these potential changes are unlikely to result in a radical alteration to the results presented here.

The next step is to use the inequality series, together with newly constructed indicators of resource dependency, to test their growth impact in a 5-year panel data regression based on a two-equation estimating framework. For the period 1900-2010 this would mean 138 observations (23*6). However, to allow comparability with previous outcomes, we may also offer results only for the XXth century (126 observations). We will adopt a similar estimation strategy as in Barro (2000) – though using a different specification. The departure point is a general framework to test for the main determinants of economic growth. Then this framework is modified or augmented to make room for additional effects operating via the new variables of interest.

15

Annex 1: Real Wage Series by Sectors

The core values for the wage series are provided by PREALC (1983).

This offers comparable data on nominal and real wages by sectors (agriculture, manufacturing and construction) as well as minimum urban wages for18 Latin American countries including our LA-6. The period covered by PRELAC ranges in most cases from 1965 to 1980. Real values are at 1970 prices in local currencies. To complete our wage series for the whole 1900-2011 period, we splice wage series from other sources to the core PREALC values by applying rate of growth backwards from circa 1965 and forward from circa 1980. In order to make the levels of the series comparable across countries we calculate PPP$ values using the PPP exchange rates available for 1970 (ECLA).

The monthly GDP per worker series are calculated as follows. First we calculate an average monthly wage from our series in 1970. Second, we use information on the wage share on total income in that year in order to calculate a GDP per worker value in 1970 consistent with our estimated average wage level. The wage rates are assumed to be 40% in Argentina, Chile, and Mexico. For Brazil we use 35% which is the average value reported for circa 1990 (IBGE), 45% for Colombia (average share of salary income on total household income during the period 1950-1970, from DNP) and for Venezuela 47% from Baptista (1997). Then we use the growth rate of the original GDP per worker (at 1970 prices) to go back to 1900 and forward to 2011. Next we present the charts by country, showing four series: real wages in agriculture, manufacturing and urban sector, as well as the series of GDP per worker. This is followed by detailed information by country on methodology and sources.

16

Charts 4a: real wage in agriculture in the LA-6