In two previous posts on this blog, I鈥檝e discussed the issue of premature deindustrialization and some of its possible consequences. In a , I, along with co-author Bret Anderson of Southern Oregon University, explored the potential consequences of premature deindustrialization further by examining the possible connections between premature deindustrialization and the defeminization of industrial employment. Premature deindustrialization is a situation in which the shares of manufacturing value added and employment begin to shrink at per-capita income levels much lower than those of the early industrializers, along with manufacturing employment peaking at lower levels. The scarce manufacturing jobs that do remain, however, are likely to be relatively high paying jobs that countries and workers compete for. In our work, we assessed whether premature deindustrialization is a feminizing or defeminizing force in industrial employment. By examining 62 countries from 1990 to 2013, we find that premature deindustrialization is likely to amplify the male bias of industrial upgrading.

The Evolving Nature of Deindustrialization and Defeminization of Industrial Employment
In a , Joe Stiglitz stated that 鈥渢he observation is uncontroversial. Global employment in manufacturing is going down because productivity increases are exceeding increases in demand for manufactured products by a significant amount.鈥� The recent academic literature provides evidence for this trend globally; lower peaks in manufacturing employment at lower levels of GDP per capita. This trend is occurring alongside another interesting trend in industrial employment; changing patterns of women鈥檚 relative employment in industry.

The of , with , 鈥� . This trend implies that the male bias of industrial upgrading outweighs the perceived female bias of export-orientated production. At the individual level, we can speculate as to why this negative relationship might occur; for example it may be due to reduced labor cost pressures in capital-intensive production, gender biases, or lack of job training for women. At the macroeconomic level, dramatic regional differences between Asian and Latin American feminization and deindustrialization patterns suggest that the economy-wide factors that condition the link between industrial upgrading and defeminization of labor may be relevant. The question that emerges is how the timing of deindustrialization is related to the male bias of industrial upgrading.

Competitiveness and the Deindustrial Regime Type
One issue in trying to answer this question is how to measure premature deindustrialization. We argue that the competitive position of manufacturing is an important link between premature deindustrialization and defeminization. Even in cases with continued success in exports, industrial upgrading has been defeminizing. See figure 1, which illustrates the relationship between capital deepening and defeminization of employment for the period from 1990 to 2013, below.

To study the potential male bias in industrial upgrading, we use an economy鈥檚 competitive position as a proxy for the deindustrialization regime type. Specifically, we suggest that prematurely deindustrializing economies are marked as being less competitive. To get to this position, we bring together Kaldor鈥檚 work, feminist scholarship, and structuralist critiques. In the manufacturing is the engine of growth. This process is related to an economy鈥檚 competitive position as the growth of manufacturing output is constrained by demand. In the later phases of development, . In , Rodrik finds that the decline in manufacturing employment that stems from a drop in relative prices is smaller in countries with more competitive manufacturing sectors. This result strongly suggests that an economy鈥檚 competitiveness is related to the deindustrialization regime.

Global competitiveness also appears within the structuralist critique literature in a related way. In particular, modern global production networks (GPN) are the contemporary analogue of past peripheral commodity exports popularized . s, , can all lead to falling terms of trade for developing countries鈥� manufacture exports. In short, the same processes that affected primary commodity exporters in the past century appear to be a barrier for exporters of manufactured goods that find themselves on the lower rungs of GPNs.

A further argues that women in the formal sector tend to lose their job faster than men, and usually have worse access to social safety nets that can ultimately impact their relative bargaining power. Further, women鈥檚 employment may be more , .

We take a few main points from this literature. For late industrializers, relative price declines in manufacturing have outpaced relative productivity gains. For premature deindustrializers, however, the decline in manufacturing employment that stems from a drop in relative prices is smaller for more competitive manufacturing countries. Finally, since women鈥檚 employment is more responsive to output fluctuations, a drop in the relative prices of manufacturing would result in a drop in women鈥檚 share of manufacturing employment relative to men鈥檚. Our overall take-away is that premature deindustrializers are likely to be in a less competitive position and, when relative prices of manufacturing fall, this fall will result in a drop in the ratio of women to men in manufacturing employment. This process leads to a lower bargaining power for women and amplifies the male bias of industrial upgrading.

Simulating the Male Bias of Industrial Upgrading
Figure 1, above, shows the negative correlation between industrial upgrading and the relative proportion of women in industrial employment. Our aim in this section is to find if there is a difference in that relationship depending on whether a country is in the premature deindustrialization category. The sample includes sixty-two countries and covers the years from 1990 to 2013. We use a country鈥檚 competitive position as a proxy for premature deindustrialization. This allows for a continuous, rather than discrete measure. We construct a , for all countries and years in our sample, which is calculated by dividing the ratio between manufacturing and total exports for a given country and year by the ratio between world manufacturing and total exports.

The figure below shows the kernel density probability distribution of competitiveness, disagreggated into two groups by timing of peak of manufacturing employment, as measured by a seven-year moving average. The figure shows that countries that peaked in manufacturing employment before 1990 tend to be more competitive, and countries that peaked more recently tend to be uncompetitive. This result lends credence to the theory that premature deindustrialization 鈥� a phenemona prevalent in late peakers 鈥� is associated with competititivness.

Next, we ran a country fixed effects regression model to estimate the impact that industrial upgrading has on women鈥檚 relative employment share in industry. Our dependent variable was the female-to-male ratio in industrial employment, and the independent variables were K/L ratio, the Balassa Index, an interactive term between the two, and several control variables. We found a positive and significant relationship between the Balassa index of competitiveness and the ratio of female-to-male workers in industry. More competitive countries have higher relative rates of female industrial employment. The capital labor ratio also had the expected negative effect, suggesting that industrial upgrading is defeminizing. Perhaps most interesting for the purpose of our question is the interactive term, which has a positive and statistically significant effect. This result suggests that competitiveness mitigates the negative effect on female employment of capital deepening. Less competitive countries, the premature deindustrializers, have a more intense male bias of industrial upgrading.

Finally, the figure below shows a Monte-Carlo simulation built from the previously described regression. It illustrates the effect of the K/L ratio on the ratio of female-to-male employment in industry by deindustrial regime type. It provides another way of seeing that less competitive countries, the premature deindustrializers, have a more intense male bias of industrial upgrading.

Conclusion
The late industrializers are deindustrializing at earlier stages of development than their predecessors and the global trends in the gender composition of industrial employment are evolving. Starting from the premise that industrial upgrading has a male bias, we tested the hypothesis that premature deindustrialization is likely to amplify that bias.

For the empirical test and simulation, we used an economy鈥檚 competitive position as a proxy for the deindustrialization regime type. The results for sixty-two countries spanning the years from 1990 to 2013 support our hypothesis that premature deindustrialization is likely to amplify the male bias of industrial upgrading. We hope to add to the evidence as we continue work on this question going forward.

Joshua Greenstein is Assistant Professor of Economics at Hobart and William Smith Colleges.

Photo: NaZemi

Any discussion of inequality includes an implicit normative or ethical comparison of distributions; a certain distribution of some good, or of gains in that good, is acceptable or not acceptable, is better or worse, is improving or stagnating. If discussions of inequality also inevitably involve rankings and comparisons of different distributions, then how inequality is . The choice of measurement of inequality is therefore not value neutral.

Relative vs. Absolute Measures of Inequality
Relative measures of income inequality are based on proportions of a total amount, or on the ratio of one group or individual鈥檚 income to another. An absolute measure depends on the actual numeric distance between groups or individuals of the unit in question. For example, suppose one individual had $10, and another $100, and then both experienced an increase of 20%, so that the first individual now has $12, and the second $120. Relative measures of inequality would register this movement as no change in distribution, since both incomes increase by the same percentage, despite the fact that, in absolute terms, the difference in income has changed from $90 to $118. It would be at least equally reasonable .

The most commonly used measures of income inequality, both in academic literature and popular discussion, tend to be relative, rather than absolute measures of inequality. The most commonly used measure of income inequality in the field of economics research for many years has been the . The gini is a purely relative measurement that assigns a number from 0 to 1, with 0 being perfect equality and 1 being perfect inequality. Other popular measures of inequality are often based on comparisons of proportions of total income along the income distribution, such as a comparison between the top of income earners sometimes referred to ,鈥� or the proportion of income of the top 1% of earners that is the focus of the research of , as well as of . These are all relative measures. (The more exotic 鈥榞eneralized entropy鈥� class of measures sometimes found in the literature, such as the , are also relative measures).

For a simple example of how the use of only relative measures can change perceptions, consider the story of developing countries, such as India, 鈥渃atching up鈥� to the US. It is quite common to hear the idea that India and China are gaining on the US economically, and from a particular perspective that is clearly what is happening; India has consistently been growing at a higher % growth rate than US, and until recent policy missteps, . Undoubtedly, this fast growth is an important phenomenon and there is a great deal of change happening within India. However, it might be surprising to many to learn, that after two decades of faster growth, the gap between India and the US鈥檚 GDP per capita has actually grown, in absolute terms. See graph 1, above.

A quick back of the envelope calculation suggests that if India鈥檚 GDP per capita were to continue to grow at 5% annually, and the US鈥檚 were to grow at 1.5.% (the approximate average from last 20 years for both), the absolute gap would not even begin to shrink for more than three decades! These average growth rates would need to continue for almost 70 more years before the two countries had an equal GDP per capita. Besides the unlikelihood of these rates continuing for that period of time, its not even clear that it is possible for all of the people in currently less wealthy countries to equal current levels of production in wealthy countries, and other . It seems fair, then, to ask in what sense Indian output per person is really 鈥渃atching up.鈥�

Right, Left, and Center
Purely absolute measurements, however, may also be misleading. As S. Subramanian of the Madras Institute of Development Studies , most would not consider a situation in which one person had $0 and another had $1 million to be equal to one in which the first person had $1 million and the second has $2 million, despite the absolute inequality being the same in both cases. A third, intermediate type of measure, explained shortly, is also a possibility.

The French economist Serge-Christophe Kolm of measures, and also equated them to different political or normative perspectives. Kolm categorized measurements of inequality as 鈥渞ightist,鈥� 鈥渃entrist,鈥� and 鈥渓eftist,鈥� depending on their treatment of inequality as an absolute or relative concept. Leftist measures are sensitive to absolute changes; they do not change when all incomes go up by the same absolute amount. An example of this type of measure would be the absolute gini, which is a standard gini coefficient multiplied by the mean of the distribution. Kolm defines centrist measures as measures which show increased inequality when average incomes rise and the relative distribution stays the same, and decreased inequality when all incomes rise by the same absolute amount. Rightist measures are purely relative; when all incomes go up in the same proportion, they are unchanged. The use, typically, of only one of these three equally legitimate types of inequality measures disguises an implicit judgment on how gains of growth should be distributed. Subramanian 鈥渢he population of a leftist perspective enables one to see that the hitherto undetected value-system underlying the relative view is rightist.鈥�

Robert Wade of LSE absolute measures are the most relevant in comparing income and wealth inequality over time in situations of growth, and would be a both a better indicator of true inequality and serve to increase the political salience of issues of distribution. Subramanian, on the other hand, on the grounds that they hold a number of desirable properties that neither of the 鈥渕ore extreme鈥� rightist or leftist can satisfy. These include an intermediate gini, which is the product of the absolute and relative gini, as well as the measurement and subsequently , which is the (a relative measure of distribution) multiplied by the (an absolute measure). Subramanian any product of an absolute and relative measure of poverty will have the characteristics of a centrist measure. Subramanian also argues that the Krtscha index is ideal for measuring changes in distribution in a situation of growth.

An Applied Example: Measuring Global Inequality
To provide an example of how using a different measure of this sort can produce a different narrative, I revisited the , who tracked income inequality globally over the last three decades. One portion of Milanovic鈥檚 work used gini coefficients to track both inequality between countries鈥� GDPs per capita, weighted by population, and the global population as a whole. I鈥檝e (roughly) recreated his results below, using data from and , as in the original.

The important point to note here is the trend of declining inequality. Now, using the same data and same two concepts of inequality, but measured using the Krtscha Index rather than a gini coefficient, the graph appears like so:

Using an centrist measure, we find global inequality between countries rising, rather than falling, throughout the 1960s, 70s, 80s, and 90s. There is a slight drop in the last decades, but inequality remains at a higher level in 2010 than it was in 1960, a very different result than that found using the gini. The interpersonal global inequality, calculated for 1988, 1993, 1998, 2003, and 2008, found a slightly downward trend using the gini, but a rising trend using the Krtscha index. Ultimately these two graphs show a very different story of what has happened to global income inequality over the last half-century.

Conclusion
At this point, a reader might object, 鈥測ou have chosen a different measure, of course you have found different results!鈥� However, that is precisely the point. The choice of measurement is determinate of a researcher鈥檚 conclusion, and the choice of measurement is not value-neutral, but indicative of an implicit judgement about what type of inequality we should care about and on how the gains of growth should ideally be distributed. If we are going to depend on relative measures virtually exclusively, then that choice should be explained and defended, not merely the .

Joshua Greenstein is Assistant Professor of Economics at Hobart and William Smith Colleges.聽

In a previous post, I wrote about the global trend of premature deindustrialization; the trend towards lower levels of industrial employment, and a shift away from industrial employment at lower levels of per capita income, and how the effects on human well-being of these trends are not yet clearly understood. An important question in understanding the impact of these changing structural patterns on individuals鈥� well-being is to whether either a lifting of the living standards of those not in formal employment, or the generation of employment to replace the manufacturing employment, is taking place.

, I illustrated how combining a household level indicator of well-being with decomposition of change analysis can shed light on these questions by focusing on two specific episodes of growth; South Africa from 1996 to 2007 and Brazil from 1991 to 2010. Using Census data from , I created indices of well-being on a scale of 0-100, using indicators such as child survival rate, access to clean water and electricity, and educations levels, culled from census data. Next, each household was assigned to a 鈥渢ype鈥� based on sectoral employment of the household head and urban/rural location, and average household scores were calculated for each type. A decomposition of change analysis was then used to assign improvement in well-being to improvement within the types and shifts in population between these types.

Findings for South Africa
In South Africa, the initially poorer performing households and groups gained more in these measurements of well-being, both in a relative and absolute sense. However, it was also the case that while there have been gains in initially poorer performing groups, large gaps still remain, and the 鈥榗atch-up鈥� has been insufficient to the extent that some groups were still performing worse in 2007 than even mid-level groups were in 1996. See Table 4 from paper, below.

The results illustrate the importance of location and employment category. The decomposition of change also illustrates that the majority of the improvement came from improvement within the lower performing groups, and much less of the improvement came from movement from lower to higher groups (See Figure 2 from paper, below). The biggest positive change from structural shift was towards low-level urban services, along with a shift towards high-level secondary jobs. The effect of this shift was positive, since the shrinking sectors were low performing types such as rural unemployed. However, households headed by low-level urban workers still trailed higher performing groups.

South Africa鈥檚 policies during this period involved big public investment in infrastructure, public services, health, education, and it certainly makes sense to attribute a great deal of the positive change in the lower performing categories to these programs. However, there was insufficient job creation during this period, and stagnation in labor intensive sectors. While there is no doubt that improved education and health among the unemployed is a positive achievement, it is also important to notice that despite these achievements, unemployed-headed households were still doing worse overall in 2007 than virtually all secondary and tertiary occupation-headed household groups were in 1996.

Findings for Brazil
In Brazil, there was less relative improvement from structural change. While in South Africa there was a shift effect from movement towards low level urban work, in Brazil, there is a large portion of the population already in such work, and the major source of improvement is within that type (See Figure 3 below). There was also a big portion of improvement explained by the improvement of urban not-in-labor market. Brazil increased its spending on transfer programs and on health and education throughout this period, and this undoubtedly helps explain these improvements.

Potential Policy Implications
One important difference from South Africa is that a higher portion of the population remained in agricultural work, and a substantial portion of the improvement also came from within that type, as well as in households headed by urban workers in low-productivity sectors. One important contrast between the two countries in terms of policy is Brazil鈥檚 supports for rural household income and rural agricultural production and South Africa鈥檚 failure to invest to the same extent in the productive capacity of the rural population. In Brazil, as in South Africa, the social programs, and, likely, rising wages for these lower level workers have led to improvement, but these groups still lag behind the better performers.

Both countries have raised the floor on living standards for the worst performing groups, certainly, but have not succeeded in truly allowing members of those groups to catch up. This result can be illustrated by the large remaining gaps between the lower and higher performing groups, and by the small to non-existent shift towards inclusion in higher performing groups. It is reasonable to suggest that despite the improvements for households in the lowest performing types, these households would be better off still if their household heads had the opportunity and ability to more productively participate in their country鈥檚 economy. These results suggest that Brazil鈥檚 recent attempts at employment generation through industrial policy may be a positive step to improve well-being. South Africa could potentially benefit from similar policies, along with some variation of the rural agricultural support introduced in Brazil.

Joshua Greenstein is Assistant Professor of Economics at Hobart and William Smith Colleges.

Photo: Jonathan Wilkins

Deindustrialization, Employment Generation, and the 鈥淧recarization鈥� of Global Labor
There are many reasons why these new patterns may have negative effects on inclusive development; emphasizes the important role that periods of high levels of manufacturing employment have played in now wealthy countries, and the dearth of wealthy countries that have skipped such a phase; there are concerns about the effects of lower levels of manufacturing output on both growth and employment generation ( See again and ).

Even aside from the direct benefits of employment in manufacturing, in both many classical growth models, such as those by Lewis and Fei-Ranis, and in , labor absorption in industrial sectors is a necessary component of eventually improving the labor outcomes for those in the rural/agriculture sectors. If labor instead moves from agriculture into 鈥渓ow-value鈥� services and the informal sector, this may not create the same opportunities for stable and growing incomes or other improvements in living standards.

Additionally, these trends have contributed to what has been described as the 鈥減recarization鈥� of labor, a declining employment population ratio, an increase in service employment, a high share of workers in the informal economy, and declining wage share and greater income inequality, which has a multitude of potential political and social effects (See van der Hoeven and , and , for examples of this argument).

Welfare Effects and the Appropriate Political Response
In , the solution to this potential development issue is to increase the provision of social and community support that is not connected to formal employment, and therefore systematically improve the well-being of those who do not have formal employment. Others, such as Amsden ( and ), or , have argued that patterns of structural change that allow more people to become directly involved in the processes of growth, especially through creation of accessible and remunerative employment, will be more successfully inclusive in terms of outcomes than those that do not, even when followed by redistribution, a viewpoint that would entail a return to greater generation of formal employment as a policy goal.

A key to understanding the impact of these changing structural patterns on individual household well-being, is to determine whether either a lifting of the living standards of those not in formal employment, or the generation of employment to replace the manufacturing employment, is taking place. If gaps remain between households in lower performing groups, such as those remaining in rural agriculture or the urban or rural unemployed, have sufficiently narrowed, that less absorption into urban industrial employment is no longer relevant, or if absorption into other employment categories, such as low level tertiary employment in urban areas, has sufficiently replaced absorption into urban industry so that there is no well-being loss from the lack of urban employment growth, then the loss of industrial employment may not be affecting well-being. If, however, large gaps in standards of living remain between households in low performing groups and households in urban industrial employment, and the move towards urban industrial employment has not been sufficiently replaced, then it can be said that the lack of employment creation in industry is negatively affecting the distribution of gains from growth.

One current focus of my own research is identifying the effects of the global trend towards premature deindustrialzaton on human well-being; a necessary in formulating the appropriate policy response. In the next blog post I will introduce my work on Brazil and South Africa.

Joshua Greenstein is Assistant Professor of Economics at Hobart and William Smith Colleges.

Photo:聽 Robert Scoble