In the mid-1990s, activists took aim at the low pay of sweatshop footwear and apparel producers for companies like Nike, many of which were located in Indonesia. Luckily, Indonesia requires firms to fill out an annual survey reporting wages, worker characteristics (education level, sex, etc.) and profits. Harrison and Scorse carefully consider, using difference-in-differences, whether activism raised worker wages, and further whether they decreased employment among exporting firms over the period 1990-1996. They find that export-oriented firms increased wages quicker, despite no major change in the characteristics (sex, experience, education, etc.) of their workers, and further than total employment did not fall among these firms. These results held particularly for low-paid unskilled workers, not for managers. Therefore, it appears that activism can positively affect third world wages by either reducing profits/skilled wages or by some sort of Card/Kreuger efficiency wage argument.
Despite the care of the work, I am skeptical. First, though employment does not fall by 1996, that’s a very short time horizon. By 2000, and certainly by 2010, there is a massive decline in export-oriented footwear and apparel production in Indonesia (much of which heads to places like Vietnam and Bangladesh with lower quality-adjusted wages). Second, the fact that export oriented apparel production wages rise from 1990 to 1996 while other wages do not, even if the authors note that there is no such divergence in the 1980s before sweatshop activism, can be implied by any number of other reasons; principally, other competitors for labor supply in SE Asia saw massive income gains from 1990-1996, often greater than that of Indonesia, and increasing wage and employment numbers in Indonesia may simply reflect a shift of apparel production to Indonesia from those countries. That is, wages may have risen for supply reasons that have nothing to do with activism. In general, issues like these two tend to be very hard to parse, hence my preference for natural experiments rather than straightforward regressions, even if the data is as good as it happens to be in this Indonesia example.
http://are.berkeley.edu/~harrison/Activism201.pdf (Link to final WP – final version in AER 2010.1)