Wednesday, August 9, 2017

Ranking Academic Economic Journals by Speed

Before, I shared some thoughts for improving academic journals. One of my complaints is about how long it takes. So, I decided to make a new ranking based on how long journals take to respond using ejmr data (among the top 50 journals ranked by citations, discounted recursive, last 10 years, although taken from last month). In the table below, # is the a journals rank by citations, and then there is data on acceptance rate, desk rejection rate, average time to first response (how the data is sorted), and then the 25th and 75th percentiles. The sample size is the last column. 

Not surprisingly, the QJE, who desk rejects 62% of papers, is the fastest, with an average time of little more than two weeks. JEEA, with a 56% desk rejection rate, is up next, followed by JHR, with a 58% rejection rate. Finance journals, who often pay for quick referee reports, tend to be fast, while Macro and econometric journals tend to be the slowest. The JME, which is notorious, clocks in at an average of 7.7 months, with a median of 4 months, but a 75th percentile of 9 months. On the other hand, the acceptance rate for those who report their submission results at ejmr is 21%, and they do not desk reject. It might be worth doing a separate ranking for those papers that actually go out to referees. In any case, among top 5 journals, the JPE is the worst, clocking in at 4.8 months on average.

Of course, when submitting, it's still journal quality/citations that matter most. But review times in excess of 6 to 9 months can be career killers. Probably, it's the right tails which are the most important here, which is why I sorted by average rather than median. 

#Journal NameAccept %Desk Reject %Avg. TimeMedian Time25th Percent. (Months)75th Percent. (Months)N =
1Quarterly Journal of Economics1%62%0.600171
12Journal of the European Economic Association4%56%1.20.50225
14Journal of Human Resources18%58%1.310238
6American Economic Journal: Applied Economics3%33%1.6202.536
31European Economic Review29%50%1.610334
15Review of Financial Studies15%20%1.821220
13American Economic Journal: Economic Policy7%40%1.820330
42IZA Journal of Labor Economics100%0%2.02221
39Journal of Financial and Quantitative Analysis33%7%2.121315
44Journal of Law and Economics0%42%2.120.5312
17Journal of Economic Growth0%0%2.12237
27Journal of Financial Intermediation14%29%2.13137
7Journal of Finance0%32%2.220419
16Economic Journal20%46%2.22.50441
19Journal of Financial Economics29%21%2.221314
34Journal of Health Economics14%57%2.320.5428
38Journal of Population Economics29%43%2.33047
5American Economic Journal: Macroeconomics11%26%2.320419
32Theoretical Economics8%0%2.3222.512
49Econometrics Journal0%60%2.40055
9American Economic Review7%45%2.420471
45Review of Finance0%9%2.532311
26Journal of Urban Economics19%19%2.531416
35Labour Economics12%35%2.521317
24Journal of Applied Econometrics0%42%2.620519
30American Economic Journal: Microeconomics23%15%2.932413
3Review of Economic Studies3%37%3.130563
21Journal of Public Economics8%25%3.231451
36World Bank Economic Review0%50%3.431.568
11Journal of Labor Economics0%20%3.542615
43Journal of Risk and Uncertainty0%75%3.83.534.54
10Review of Economics and Statistics2%50%3.820650
23Journal of Development Economics14%33%3.831536
25Journal of Business and Economic Statistics22%11%3.93259
41Journal of Environmental Economics and Management24%10%4.03.52521
22RAND Journal of Economics7%15%4.144527
47Journal of Economic Dynamics and Control50%17%4.231418
33Journal of Economic Theory17%13%4.2435.523
46Journal of International Money and Finance41%6%4.432617
50Oxford Bulletin of Economics and Statistics23%38%4.642613
2Journal of Political Economy0%60%4.831.5725
18Journal of International Economics19%6%
20Journal of Money, Credit, and Banking20%15%5.142820
40Journal of Economic Surveys0%40%5.25075
28Experimental Economics29%14%5.96397
29Journal of Econometrics11%11%7.075109
48Econometric Theory0%0%7.054123
8Journal of Monetary Economics21%0%7.743914

Sunday, July 30, 2017

The Saga Continues: A New Addition to the Currency Unions and Trade Literature

Previously on this blog, I have written about the saga of the Currency Unions and Trade literature. This literature began with Andrew Rose, the famed discoverer of the finding that currency unions, like the Euro, appear to have an effect on trade that is nothing short of miraculous. Effect estimates range in the 100% to 1,300% range, according to researchers at places like Harvard and Berkeley.

I published my very first academic paper about this topic, and found that the earlier large estimates of currency unions (CUs) on trade were driven by rather blunt omitted variables, such as warfare, decolonization, and communist takeovers, and were also sensitive to dynamic controls. I wrote that countries joining the Euro should not expect any large effect on trade.

A new paper by Glick and Rose came out last year which used more recent data, and, once again, found a large impact of CUs, including for the Euro. I was, once again, skeptical, so I assigned my undergraduates a search-and-destroy mission. This was aided in part by Andrew Rose's very laudable practice of posting his data online, which allowed my students to search and destroy. The original authors, to their credit, responded in the comment section of that post. I posted Reuven Glick's thoughtful response here, along with my own response.

In any case, Aleksandr Chentsov and I decided to go ahead and write up a new paper on the topic: "Breaking Badly: The Currency Union Effect on Trade". In the paper, we essentially tested whether these same omitted variables which were driving the effect initially were also driving the effect using this much larger dataset, and whether omitted variables (think the EU) might also be driving the results for the Euro Area as well.

The basic problem can be seen from the evolution of trade between Pakistan and India (Figure 1 below). After the dissolution of the currency union in 1965, trade did, in fact, plummet. By 99.8%. It would thus seem to provide evidence for a large impact of CUs on trade. If Greece leaves the Euro, one might wonder that something similar might happen. However, it doesn't exactly take an expert in International Relations to know that India and Pakistan haven't always gotten along swimmingly. 1965 also happened to be the year when a brutal border war broke out over the legacy of partition, following Pakistan's "Operation Gibraltar". It provides a better guide to what might happen if Greece defaults on all of its debts, gets kicked out of the Euro Zone, and then the EU invades it in retaliation, but Greece fights it to a stalemate.

Figure 1: Trade Between India and Pakistan. The vertical red line shows the end of the Currency Union, which also happened to coincide with the Indo-Pakistani War of 1965.

The example of India and Pakistan was hardly an isolated case. We write in the paper that "In addition, all of the countries which left the French Franc -- Tunisia, Algeria, and Morocco -- did so after major conflicts resulting in independence (see Thom, 2006). These included separatist bombings in the case of Tunisia, a war of independence in the Algerian case, and anti-colonial rioting in Morocco. All five of Portugal's former colonies which had also shared currency unions likewise had to fight for their independence, some of which included prolonged guerrilla wars." When you exclude these cases, the measured CU effect shrinks.

This brings us to the Euro. Just as leaving a currency union -- which, like marriage, are meant to be forever -- is typically a sign of geopolitical turmoil, joining a currency union is typically a sign of good/improved/improving political relations. In the Europe case, my students noted that one would want to control for the entire history of European integration, from the Coal and Steel Community, to the EU. One perceptive student noted that prior to the 1990s, some parts of the Euro Zone today, such as East Germany and other late joiners in Eastern Europe, were all part of the Warsaw Pact.

Our goal, then, was to find appropriate control groups for both Western and Eastern Europe. For Western Europe, we used either (1) other EU countries not in the Euro, or (2) other Western European countries not in the Euro. For Eastern Europe, we compared the evolution of trade between the EE Euro entrants and other EE countries not in the Euro. The results in either case did not suggest a measurable/significant impact on trade. In Figure 2, we plot the evolution of trade in Euro Area countries in Western Europe relative to trade with Non-Euro countries. Relative to 1999, we actually found that Euro members traded slightly more with non-Euro members in 2013, although the difference wasn't even close to statistical significance. However, Euro members had experienced a dramatic increase in trade in the 1950s. Thus, a simple dummy strategy which averages trade before and after the Euro was formed can lead one astray.
Figure 2: Trade intensity of Euro Members relative to Non-Members over time. The vertical red line in 1999 denotes the formation of the Euro. 

We do much more in our paper. Looking at each major CU separately, we find that there really aren't any clear-cut examples for the CU effect, but there are many counterexamples (such as the Euro above). Often, there were dramatic trade declines in the final years of a CU, and then trade recoveries after dissolution. In addition, we also ran the traditional panel gravity regressions, and once again found that the results are sensitive to omitting the CU switches which coincided with war or other major geopolitical events, and adding in other CU-specific controls (such as for the EU).

Some general lessons for empirical research in international trade that research on this topic taught me include:

(1) one should always be mindful of dynamics. Particularly when regressing a level variable that trends on another variable which trends. (Most country pairs have just one CU switch, so any trend in the data could lead to incorrect inference.) This is often my first or second concern when I see papers presented at conferences.

(2) One should always plot their data. I think many authors do not do enough of this. Doing so can allieve the first concern. In the more recent version, Glick and Rose did at least plot pre-treatment trends, a clear improvement. But the existence of pre-treatment trends implies a non-randomness of the treatment.

(3) Are the errors clustered appropriately? In this paper we found we could shrink the t-score on the Euro impact by 80% simply by using multi-way clustered errors.

(4) Is the effect size plausible based on what else we know? In this case, we knew that currency pegs are correlated with much smaller effect sizes, and that indirect pegs -- more likely to be random -- are not correlated with higher trade flows at all. In addition, the effect sizes which have been bandied about in this literature were orders of magnitude larger than, say, the Smoot-Hawley tariff. Simply too large to be believed. And Glick and Rose also had argued that some CUs cause sharp contractions in trade, while others had no effect, even while others had very large positive effects. Why such dramatically different effects for each CU? The answer is that there were simply different historical forces in play for each CU, and these forces overshadowed whatever small effect CUs may have.

(5) Always think about endogeneity! It's such an obvious, and ubiquitous problem that I don't know if it's necessary to add this point. But I do think this deserves to be a textbook cautionary tale of a non-random treatment leading researchers badly astray.

Is this research that important in the end? Admittedly, most countries that joined the Euro did not do so based on their belief of the CUs and trade literature. Nevertheless, the Euro has, in my view, been mostly a catastrophe for southern Europe. I believe the first best option for these countries would be more aggressive pro-growth stimulus from the ECB, but, absent that, I think these countries should think seriously about exit. While the Euro is a bit different from most other currency unions (the definition is that two countries have currencies that trade at a 1:1 par value), there is no hard evidence that Euro Area countries will face a trade collapse if they leave the Euro.

In any case, I did have a senior economist sit me down and tell me not to write this paper. The logic is obvious. You get places in academia by forming close relationships with powerful people, not by pissing them off. These guys no doubt have close relationships with many other economists in the field. They likely also referee lots of papers a year, and will likely be asked to referee this paper. Most editors understandably won't want to touch this controversy with a 10-foot pole, (several websites took a pass on a column about this paper, one on the grounds that my coauthor and myself are at less prestigious institutions than Glick and Rose; how could an MIT Ph.D. be wrong?) while many potential referees are also no doubt close friends with the authors. I expect to submit this paper 7-10 times, but that is relatively normal. The original researchers will no doubt take my criticism of their research personally, and will likely do everything in their power to sabotage my career. Undoubtedly, there are powerful corrupting forces in academia.

Or, maybe not. Maybe the authors will understand that it wasn't personal. Maybe the editor and referees will judge our paper based on its merits. The only way to find out is to write the paper and submit it. So that's what we did.