Paul Walker

Dr Paul Walker is an economist at University of Canterbury. He has expertise in microeconomics, institutional economics and industrial Organization. He blogs for The Dismal Science.

The Economics of World War I. 3 - The Dismal Science

Sep 13, 2014

Another in the series of posts from The Economics of World War I project at

Walking wounded: The British economy in the aftermath of World War I
Nicholas Crafts 27 August 2014
It is well-known that World War I was expensive for Britain. The indirect economic costs were also huge. This column argues that the adverse implications of the Great War for post-war unemployment and trade – together with the legacy of a greatly increased national debt – significantly reduced the level of real GDP throughout the 1920s. A ballpark calculation suggests the loss of GDP during this period roughly doubled the total costs of the war to Britain.

Even sweatshop jobs can be good for you, - The Dismal Science

Sep 12, 2014

especially if you are female.

A new NBER working paper looks at Manufacturing Growth and the Lives of Bangladeshi Women. The paper is by Rachel Heath and A. Mushfiq Mobarak.

Heath and Mobarak note that there is very little rigorous analysis of the welfare effects of access to factory jobs - "sweatshop jobs" - for women, and the "evidence" that there is is dominated by anecdotes from anti-sweatshop activists about the negative effects of hazardous working conditions. The evidence gap is especially pertinent for policymaking in Bangladesh, where well-publicized garment factory collapses and fires and their attendant death tolls recently captured the world’s attention. In response to the disasters individual large buyers, in addition to the U.S. government, have made moves to restrict or boycott garment exports from Bangladesh but such export restrictions and boycotts come with the potential to hurt the very workers that they are designed to protect.

The manufacturing industry in Bangladesh currently employs almost 4 million workers and it was the first industry to provide employment opportunities to women on a large-scale in a country where women traditionally have not worked outside the home. In doing so to it has raised the opportunity cost of being married and having children. These sweatshop jobs require a basic level of literacy and numeracy skills and the arrival of garment factories therefore has potentially affected enrollment, employment, marriage and childbearing decisions for Bangladeshi women.

Heath and Mobarak's

[...] results suggest that the rise of the garment industry can help explain the declining fertility, increasing age at marriage, and rapid increase in girls' educational attainment during this period, both in absolute terms and relative to boys [...]. Approximately 80 percent of garment factory workers in Bangladesh are female [...] and extrapolation from our data and national surveys suggest that around fifteen percent of women nationwide between the ages of 16 and 30 work in the garment industry. The education results are particularly policy- relevant, as Bangladesh has surpassed the third Millennium Development Goal of gender equity in enrollments, a goal with which many other countries in Western Asia and sub-Saharan Africa continue to struggle. Our research design permits a study of investments in girls relative to boys, which is of considerable policy [...] and also academic interest, given the comparative advantage girls possess in skilled tasks [...]. Our results provide one hitherto unexplored explanation for the accelerated gender equity in education in Bangladesh, thus generating important policy implications for other developing countries interested in emulating Bangladesh's success.
There are, in theory, a number of different ways that access to factory jobs can alter women’s school, work, marriage and childbearing decisions.
As documented for maquiladora jobs in Mexico, older children may be induced to drop out of school to access the factory jobs [...]. Conversely, younger children (who are still too young for the factory jobs and do not face the temptation to drop out and begin earning immediately) may respond by investing in education if the availability of relatively well-paid manufacturing jobs increase the returns to education. Educational attainment might also increase through a wealth effect, if parents can now better afford to send their children to school. Both the increased enrollment channel and the direct factory employment effects would result in girls delaying marriage and childbirth.
Heath and Mobarak
[...] document that the hazard of marriage and childbirth at early ages (12-18) drops sharply for girls when they gained exposure to the ready- made garment sector. This is important because other research has documented large negative welfare implications of early marriage and early childbirth [...].
Next Heath and Mobark examine the ways by which these delays in marriage and childbirth occur.
Did girls increase their educational attainment in order to obtain well-paying garment jobs which require numeracy and literacy, which then led to a postponement of marriage due to greater educational attainment? We assess the effect of cumulative years of exposure to garment factory jobs on the total years of educational attainment (for those with completed schooling histories), adding an additional comparison to the girls’ male siblings, given that garment production is has been a much larger innovation in the labor market for girls than for boys [...]. We find that girls gain an extra 1.5 years of education relative to their brothers in the median garment-proximate village. This represents a 50% increase in girls’ educational attainment over control villages that do not have a garment factory nearby. We observe the increase in female education (relative to their male siblings) even in families where the mother or older sister never worked in a factory, which suggests that increased demand for skills in factories that offer job opportunities for women is a likely channel through which the enrollment gains are realized, in addition to any family wealth effect or changes in intra-household time allocation from other household members working in garment factories.

We next use retrospective data on the entire history of annual school enrollment decisions for each girl to test whether the effects of the garment industry on schooling are strongest for younger girls. We find that young girls (aged 5-9) are more likely to stay enrolled in school hen factories open close to their village compared to girls in comparison villages in the same sub-district that are not located within commuting distance of factories, relative to earlier years (before the factory opens), and relative to male siblings of the same age.
While there is a positive effect on education for young girls, what about those a bit older?
Our data also indicate that the delays in marriage and childbirth we estimate likely also stem from girls in garments-proximate villages choosing to work in factories when they are about 17-23 years old, instead of getting married (or staying in school). Factory job access has a small negative effect on school enrollment of 17-18 year olds (unlike the positive effect for younger girls). More importantly, girls who are exposed to factory jobs when they are between 10 and 23 years old (which is the critical age group at risk for early marriage in Bangladesh) are 17 percentage points more likely to have done wage work outside the home, and this is a 79% increase over the control group.
In summary,
[...] access to factory jobs significantly lowers the risk of early marriage and childbirth for girls in Bangladesh, and this is due to both the girls postponing marriage to work in factories, and the girls staying in school at earlier ages. We benchmark the magnitude of the effects of the garment industry against the effects of a large-scale (US$15 million per year) conditional cash transfer for schooling intervention run by the Bangladesh government with multilateral donor support. The program has paid for 2 million girls to remain in school, conditional on remaining unmarried. The dramatic improvement in girls' outcomes in Bangladesh in the past 30 years has 5 frequently but casually been attributed to the FSP, but our estimates suggest that the rapid expansion of the garment sector has been a much more important reason for the decreases in earlier marriages and fertility and the closing of the gender enrollment gap in Bangladesh.
While sweatshop jobs are not great jobs, by our standards, they do come with positives for those people in poor countries who can get them. The anti-sweatshop groups should keep this in mind when attacking such forms of employment. Sweatshops are the first step down the road out of poverty.

Econtalk last week - The Dismal Science

Sep 10, 2014

Nathan Blecharczyk, co-founder and chief technology officer of Airbnb, talks with EconTalk host Russ Roberts about Airbnb, one of the earliest companies to use technology to allow individuals to share underused resources, and in the case of Airbnb, hou...

Do economists play well with others? - The Dismal Science

Aug 30, 2014

The unlikely sounding answer to this question is yes. This answer comes from a paper in a recent issue of The American Economist (Vol. 59, No. 1 (Spring 2014)) on "Do Economists Play Well with Others? Experimental Evidence on the Relationship Between E...

The effects of patent trolls - The Dismal Science

Aug 28, 2014

There is a new organisational form, called the non-practicing entity (NPE), in the world of intellectual property. NPEs have recently emerged as a major driver of IP litigation. The idea is that NPEs amass patents not for the sake of producing any actual product, but rather they aim to prosecute infringements of their patent portfolios. (rent-seeking?) The rise of NPEs has sparked a debate regarding their value and their impact on innovation. Proponents argue that imperfections in the legal system implicitly reward large, well-funded organisations, enabling them to infringe at will on small innovators’ IP and that NPEs are there to protect small innovators from such abuse. Opponents cast NPEs as organisations that simply raise the costs of innovation by exploiting the fact that an imperfect legal system will rule in their favour sufficiently often—even if no infringement has actually occurred—that the credible threat of the legal process can yield rents from producing, innovative firms.

So what are the effects of these "patent trolls"? A new NBER working paper, Patent Trolls: Evidence from Targeted Firms by Lauren Cohen, Umit Gurun and Scott Duke Kominers, tries to find out. Cohen, Gurun and Kominers add to the debate on NPEs by providing the first large-sample evidence on precisely which corporations NPEs target in litigation, when NPE litigation occurs, and the impact of NPE litigation on the targeted firms’ innovative activity.

Cohen, Gurun and Kominers argue that there are two reasons that patent trolls can prevent welfare-increasing innovation from being brought to market.

  1. innovators with profitably commercialisable inventions but with a high enough probability of being sued to be deterred from production
  2. innovators that decide not to commercialise because the ex ante expected profitability of becoming a patent troll is higher than that of commercialisation
In their empirical work Cohen, Gurun and Kominers
[...] link patent-level data on NPEs and their activities to data on all publicly traded firms. Using this linked data, we show that NPEs behave opportunistically; that is, typically acting as patent trolls. Specifically: NPEs target firms that are flush with cash (controlling for all other characteristics) and firms that have had recent, positive cash shocks.

Indeed, a one standard-deviation increase in cash level increases the probability of being sued by an NPE by 11% (t = 6.84). Given that the mean probability is 2%, this is more than a fivefold increase.

In fact, NPEs even target conglomerate firms that earn all of their cash from segments having nothing to do with the allegedly infringing patents. For example, an NPE is likely sue a firm regarding a technology patent even if the firm is earning all its revenue from a lumber division entirely unrelated to the allegedly infringing technology patent—even if the division holding that patent is unprofitable. Indeed, we find that profitability in unrelated businesses is almost as predictive of NPE infringement lawsuits as is profitability in the segment related to the allegedly infringing patent.

Consistent with our model, we also find that NPEs target firms against which they have a higher ex ante likelihood of winning. We demonstrate this fact using multiple measures of ex ante likelihood of lawsuit success. First, we show that NPEs are significantly more likely to target firms that are busy dealing with a number of other litigation events unrelated to intellectual property. Being tied up with outside litigation roughly doubles the probability (t = 2.87) of being sued by an NPE. Moreover, we show that, controlling for all other characteristics, firms with larger legal teams have a significantly lower probability of being targeted by NPEs, consistent with large legal teams serving as a deterrent.

Of course, the true prediction of our model is on the ex ante expected profitability of NPE litigation. To capture this, we interact our measures of expected cash payouts with our measures of expected lawsuit success. We find that, as the model predicts, NPEs systematically target those firms for which the ex ante expected profitability of litigation is large. In particular, the payout probability interaction terms are significant and economically large. Our finding suggests that nearly all the firms targeted by NPEs have large pools of cash for potential payouts and are ex ante more likely to pay off in some form (either an out-of-court settlement or an in-court loss). To further explore this connection, we construct a measure of the ex ante expected outcome if a targeted firm were to go to court. This measure relies on the assumption that defendants often make predictions about the likely outcome based on observations of other firms in the same industry and location. We find that the interaction term of this expected outcome and expected payout is again large and significant, providing further evidence that NPEs choose targets based on expected profitability: suits with high probability of payoff against firms with deep pockets.

Non-practicing entities don’t have a monopoly on IP litigation. Practicing entities (PEs), such as IBM and Intel, also sue each other for patent infringement. If our results are simply picking up general characteristics of IP litigation, then we might expect to see PEs behaving in much the same way as NPEs. In order to compare PE and NPE behavior, we hand-collected the universe of patent infringement cases brought by PEs against other PEs in the same period (2001–2011). However, we find the opposite. If anything, PEs are slightly less likely to sue firms with high cash balances and less likely to sue firms with many ongoing cases. All of the other determinants of NPE targeting have (statistically and economically) no impact on PE litigation behavior. This comparison suggests that our results on NPE litigation behavior are not just reflections of general characteristics of IP litigation. Rather, our findings are consistent with agent-specific motivations for NPEs in targeting firms flush with cash just when favorable legal outcomes are more likely.

Lastly, we examine the real impacts of NPEs’ litigation activity. Comparing firms that are sued by NPEs and go to court (and in this way controlling for selection of firms targeted by NPEs), we find that firms that lose in court have significantly lower post-litigation patenting activity and fewer citations to their marginal post-litigation patents, relative to firms whose cases are dismissed. Furthermore, after losing to NPEs, firms significantly reduce R&D spending—both projects inside the firm and acquiring innovative R&D projects outside the firm. Our evidence suggests that it really is the NPE litigation event that causes this decrease in innovation. Prior to litigation, firms that subsequently lose to NPEs are identical to those that subsequently have suits dismissed. They have the same R&D, patenting, and patent quality. Moreover, patents of firms developed pre-litigation continue to accrue citations at exactly the same rate after litigation, whether or not the suit was dismissed. This is in stark contrast to the divergent amount of citations of firms’ post-litigation patents.
In short, NPEs behave as patent trolls.

Should change the way we teach economics? - The Dismal Science

Aug 27, 2014

This is a question that Professor, and Nobel Prize winner, Alvin Roth was asked by a reporter from Brazil. The questions and Roth's answers follow:

1) Should the content of economics degrees change? Why? Why not?

I guess you mean should we change what we teach young economists, and of course the only answer is “of course!” What we teach young physicists and biologists and doctors and civil engineers changes as we learn more about those things, and economics is no different.

2) Has the criticism of economics been exaggerated after the 2008 crisis? To what extent is the current debate on content useful?

I think the 2008 crisis has been useful for pointing out that economics is, in many of its parts, still a very young science. For an analogy, think of medicine, which is the part of biology that we most often look to for advice, and is also a young science in many of its parts. Each year we worry that there might be an influenza epidemic due to whatever new strain of flu is observed in Asia that year, and each year vaccines are prepared, in an attempt to avert a disaster like the influenza pandemic of 1918. So far we've been lucky, but it’s not because we have a deep understanding of what could cause another epidemic or how to prevent it. But if another epidemic occurs, we’ll need to rely more on doctors and medicine, not less. So, while we need to understand epidemics better, that’s not a deep criticism of medicine, just an acknowledgement of some of its current limitations. Similarly for economic crises, and economics.

3) What changes should be made?

One of the things we’re devoting more attention to at Stanford is the kind of economic engineering called market design, which pays attention to the detailed rules by which particular marketplaces operate, and to experimental economics, which gives us a tool to better understand how people behave in economic environments.

4) Has economics teaching become too wedded to scientific pretension? Was excessive faith invested in abstract mathematical models?

Abstract mathematical models are very useful, in combination with other kinds of investigation. A lot of my work is devoted to market design, and my colleagues and I build a lot of marketplaces that have some of their ancestry in abstract mathematical models (including some of those explored by the famous Brazilian economist Marilda Sotomayor, in whose honor there is a conference next week). Mathematical models are becoming increasingly important as we start to explore really big data sets, since not only do you need mathematical tools to test hypotheses on data, you need models to even suggest what hypotheses you should be testing. Theory and observation work best in combination…
On this last point one of the top economic theorists in the area of the theory of the firm and contract theory, Oliver Hart, has noted (and this would be my answer to Matt Nolan's recent Discussion Tuesday question),
Although theory may not be as prominent as it once was, it remains essential for understanding the (increasingly) complex world we live in. One cannot analyze the bewildering amount of data now available without the organizing framework that theory provides. I would also suggest that one cannot understand the extraordinary events that we have recently witnessed, such as the financial crisis, or make sensible policy recommendations in response to these events, without the organizing framework of theory.
So for those who seem to think data can do everything, and we should therefore stop teaching theory, I don't think so. Empirical work is only as good as the theory underlying it. So, no, running a million regressions and picking the one that confirms your prejudices isn't how you do good economics.