Wednesday, January 29, 2014

Voting in a Globalized Economy: Lessons for (and from) Turkey?

The major news to come out of Turkey last night was the decision by the Central Bank to raise interest rates substantially in order to prevent further decline in the value of the Turkish Lira. How well that will work is another matter, but it highlights a crucial tension that governments in Turkey (and most states) face today: how to balance the demands of global market pressures with those of domestic constituencies.

The challenge is this: raising interest rates will tend to drive up the exchange rate of the currency. However, it will also tend to slow down the domestic economy by pushing up the costs of credit and borrowing. For this reason, Turkish Prime Minister Recep Tayyip Erdogan had been reluctant to do so prior to this year's local and Presidential election, given the well-documented relationship between the state of the national economy and election results (see this for a lot of citations).

The increasing globalization of many states over the past 30 years has led to a different concern. As states have become more reliant on international investment to finance and sustain development, they have put their own economies--and hence their economic policies--at the mercy of impersonal global markets. The concern this raises is whether democracy is diminished (over even sustainable) under such conditions. The problem is like this: if a leader would like to pursue a certain economic policy, and s/he was elected with majority support to do so, but global market forces "demand" the opposite policy, then the leader must either defy the will of his/her electorate or risk market punishment. This possibility led economist Dani Rodrik to suggest that there is a "political trilemma" between national sovereignty, globalization, and democratic politics. Under this trilemma, we can have two of the three, but not all three simultaneously and all the time.

Elections under Globalization?
Drawing on this logic, scholars of elections have debated how these function in a global economy. This has been an interesting debate, because several possible results have been suggested.

To review: the standard argument of economic voting is that electorates reward and punish incumbent governments for their management of the national economy. When times are good, incumbents win re-election. When the economy is bad, voters "throw the bums out" of office.

But what if voters know that they live in a globalized economy? Two problems emerge:
  1. National leaders have to balance the desires of their citizens with the demands of impersonal global market forces.
  2. Stuff happens. A global financial crisis, or other economic developments elsewhere (such as the Fed's tapering) may cause economic repercussions in your country that your own economic policies cannot prevent. 
If voters understand these two things to be true, then how will they respond? Three arguments exist.

Government Constraint Hypothesis
Best associated with the work of Tim Hellwig and co-authors (gated), this argument suggests that economic voting will end. Because voters understand that their leaders cannot do very much to influence economic conditions in their country, they will choose to base their vote on factors other than the economy. This would suggest the end of economic voting, and we should less evidence of a correspondence between the state of the economy and election results as countries become more globalized.

Benchmarking Hypothesis
Best expressed by Mark Kayser and Michael Peress, this hypothesis argues that globalization actually provides opportunities for voters to improve their assessments of economic performance. In short, the argument goes like this: global economic trends affect all economies, so relative over- or under-performance compared to similar national economies tells voters that their own government is doing a better or worse job of managing the economy. In other words, if your economy is growing 2% per year, but your neighbor's economy is growing 3% a year, then your government must be doing a bad job and you should vote them out of office.

Null Hypothesis
There are various ways to arrive at the hypothesis that globalization does not change economic voting. One might be simply that voters are generally too unsophisticated to figure out that their governments have lost the capacity to manage the economy under globalization. This would fit with the work of Larry Bartels, who has argued that voters are generally pretty short-sighted and unsophisticated when it comes to holding governments accountable in a book and series of papers with Chris Achen. (Note: I have a chapter in a book manuscript under review that argues this point with regards to the Global Financial Crisis. Hellwig and Kayser/Peress also have chapters in that book).

(Another possibility, suggested by Andrew Healy and Gabriel Lenz, is that myopic news reporting of economic performance leads voters to focus on recent events).

Evaluating Turkey's Elections?
So what can the upcoming Turkish elections tell us about all of this? One might be tempted to feel sympathy for the governing Justice and Development Party (AKP), seeing that higher interest rates may slow down the Turkish economy. However, it's important to remember that the AKP benefited in 2007 and 2011 from a fast-growing economy--one that was aided by a conducive environment for developing economies like Turkey (Rodrik makes the point here that Turkey's growth record from 2002-12 was unremarkable by the standards of similar economies).

In my previous post, I set up a (very) simple model of economic voting, which shows a reasonably strong correspondence between the state of Turkey's economy and the AKP share of the vote. Based on current economic predictions (as of a few days ago, anyway), the model would predict the AKP to receive about 43% of the vote in the March local elections. The model indicates that each 1 percentage point in real GDP growth raises the incumbent's vote share by 0.84 percentage points, suggesting a strong relationship.

All of this suggests that Turkish voters are neither discounting nor benchmarking the economic record of the Erdogan government. I think this makes sense. It may not, strictly speaking, be due to a lack of voter sophistication or information. Rather, it may owe to heavy doses of motivated reasoning among partisans of various stripes. AKP supporters will credit the government for good times and rationalize bad times to factors beyond its control (or conspiracies involving a global "interest rate lobby"), and vice versa. The relatively smaller group of unattached voters turns out in practice to include many who are less interested in, and thus knowledgeable about, politics. Precisely the sort of voters who are unlikely to engage in discounting or benchmarking.

So it will be interesting to see how voters respond to the economic environment in these upcoming elections. (It will be interesting to see how the economy is by the time of these elections!). Do they punish the AKP for a slowdown, if in fact one happens? Or do they let the Erdogan government off easy due to the fact the pressure on the Turkish Lira is due, in part, to the US Fed's "taper"? Whatever happens, this will tell us a little bit more about how voters respond to economic changes in an open economy.

Monday, January 27, 2014

Projecting Turkey's 2014 Local Elections

In March, Turks will go to the polls to vote in local elections. Given all of the events in the past year, there is a great deal of interest: will voters punish Prime Minister Erdogan and his AKP for its authoritarian turn and revelations of corruption, or will voters in fact ratify his support? Given the historic instability of Turkish democracy and the less-developed nature of public opinion polling, projecting the results of a Turkish election is trickier than in the US or Western Europe. Nonetheless, I will join a few others in offering a prediction of sorts.

Thinking About Elections
Some commentators have already offered their opinions. Steven Cook of the Council on Foreign Relations has predicted that Erdogan's AKP will win the upcoming elections, arguing on the basis of his expert analysis that Erdogan has outmaneuvered his political rivals and that the opposition CHP has little to offer voters. Columnist Tulin Daloglu, after interviewing people in several AKP-leaning Ankara neighborhoods, argues that Erdogan remains popular.

All of these predictions are fine, but they have their problems. Cook's claim is pretty hard to pin down. Saying that AKP will win the overall vote is a bit vague, since I can't think of anybody who would predict otherwise. Daloglu's commentary would be more insightful if she had found evidence of AKP supporters who were swaying. But, as it is, she basically is able to report that AKP's core voters remain loyal. This isn't terribly interesting, as we would expect that they would be the last to fall out with Erdogan. (On the other hand, some of the quotes she gathers make for fantastic examples of motivated reasoning among Erdogan loyalists, so well worth the read).

One problem many political journalists and pundits have in projecting elections is that they forget two important facts about voters:
1. Voters think about politics far less than experts do.
2. Most voters vote for the same party/ideological family every time, even if they insist otherwise. True "independents" or "swing voters" are far less common than many pundits imagine.

Thus, too much commentary on elections focuses on big events, speeches, or scandals--imagining that voters will respond to these events the way that political elites. I saw one person claiming on Twitter that Kemal Kilicdaroglu (the CHP leader) had destroyed his party's chances of winning by giving an uninspiring speech on 17 December, when the corruption investigations were launched. This really misunderstands how voters approach politics and elections. Why? Many voters don't pay that much attention, and they don't know whether Kilicdaroglu gave a speech on 17 December, much less whether it was any good. Moreover, most voters will process the new information they receive to fit their prior political beliefs. Hence, AKP voters will accept Erdogan's claims of foreign conspiracies (again, see the quotes in the Daloglu article!), while opposition supporters will view accusations of corruption as proof against Erdogan. What this means is that, if we want to get a more accurate sense of how the elections might unfold, we need to find out the attitudes of those whose support for any one party is not absolute, and we need to understand how they are actually thinking about the unfolding political events (not how we are thinking about them!). That can be tough to do for various logistical reasons, but there are ways we can still make reasonably informed predictions.

Predicting the Election Result
The general stability of many voters' behavior means that we can focus on predicting change within a relatively narrow band of outcomes. In other words, it is unlikely that AKP will win either 80% or 20% of the vote in a fair election, so we can take for granted that they are likely to get somewhere between, perhaps, 30% and 60%. Moreover, we know that voters are highly responsive to specific concerns--namely, peace and prosperity.

How can we do this? Election forecasting is a fairly big industry in advanced, stable democracies such as the United States or the United Kingdom. There are a variety of approaches, but the most interesting are those that rely on very little information and can thus make a reasonable prediction months in advance. Such forecasting models typically rely on some measure of the national economy (or voters' perceptions of the national economy), along with 1-2 simple political variables. A good example is Alan Abramowitz's "Time for Change" model, which predicts US presidential elections on the basis of real GDP growth, presidential approval rating, and the number of terms the incumbent party has held the White House. With only three variables, he can still achieve high predictive accuracy. Why? Because, again, we understand the high correlation between the state of the national economy and election results. "It's the economy, stupid!" may be a cliche, but it's very true when it comes to elections.

Projecting the outcome of the Turkish election becomes even more difficult, as there have been fewer elections under the current party system (put simply, it is hard to figure out how elections in the 1990s DYP/ANAP/FP/DSP era correspond to today's party system) and there is less polling information on things like leader approval. So, I'm going to go extremely (arguably, ridiculously) basic, and project the AKP vote share on the basis of nothing more than real GDP growth. What's more, I've only got four elections--the 2 general elections and 2 local elections since 2002--to work with. So...take this with a heavy dose of skepticism. (And, if anybody can point me to reliable information with which I could improve this model, I'd love to see it!).

A brief technical aside: I played around with different measures of the economy. The measure that fitted the data best was a weighted measure of the election year's real GDP growth rate and the previous year's real GDP growth rate, with the latter being weighted 50% of the former (source: World Bank).

So...here goes:


While a projection based on only four data points stretches credibility, it is worth noting that the data paint a pretty consistent picture (albeit, one that could be shattered by the next round of elections). The slope of 0.84 implies a fairly meaningful relationship between real GDP growth and the AKP's vote share as incumbents: the AKP's best result (just under 50% in 2011) came during a two-year period of booming growth of around 7% per year.

What does this project for the 2014 elections? Further caution is in order, since the economic data for 2013 are estimates and projections for 2014. Still, the World Bank claims that Turkey's real GDP grew by 3.6% in 2013 and will grow by 3.8% in 2014. That yields a prediction that the AKP will receive 43.5% of the national vote. Note that this actually comes pretty close to a recent poll by the Turkish polling firm Sonar, which put the AKP on 42% and CHP on 29%.

More than a few caveats...
There are plenty of reasons to be cautious, of course, many of which might point to a lower estimate for the AKP:
  • The effects of the Gezi Park protests and the ongoing corruption investigations/responses. While core AKP supporters may not be defecting, it is possible that other previous AKP voters will. It is hard to estimate how much (or little) of an effect these events might have on influencing voting behavior.
  • Some pundits, such as Steven Cook, remain skeptical that many voters would turn to the CHP. But, as I've shown in past research (gated), this is not the only alternative. Supporters of the AKP may simply not turn out to vote in as high numbers as previous elections, which would have the practical effect of reducing the AKP's vote share.
  • These are local, not national elections. While local elections are typically second-order and thus largely function as referendums on the national government of the day, local concerns and personalities matter to some extent. In addition, second-order elections tend to stimulate more protest votes. It is worth noting that the AKP received only 39% of the vote in the previous round of local elections, though this was in a tougher economic environment.
  • The potential for election fraud. Will the Erdogan government allow for free elections, or will AKP officials engage in fraud to win crucial local elections in Istanbul and Ankara? Given the behavior of the Erdogan government during the Sledgehammer and Ergenekon investigations, as well its attempts to bypass the judicial process in the ongoing corruption scandals, why would we expect that they would not?
Thus, my "model" (such as it is) predicts the AKP to receive 43% of the vote. That said, I would say there are more reasons to lower that prediction (though only by perhaps 2-4 percentage points). A result similar to that of the previous local elections could well be the outcome: the AKP around 40%, with the CHP perhaps close to 30%.

It is worth remembering, of course, that the national vote totals do not matter except for their symbolic value. What is actually at stake is the control of hundreds of mayorships and local councils around the country. And, given their size and symbolic value, most eyes will be on Turkey's two largest cities--Istanbul and Ankara.

Sunday, January 12, 2014

Get Tenure or Die Trying? Understanding the Academic Job Market

A couple months ago, political economist Alexandre Afonso wrote a blog post called "How Academia Resembles a Drug Gang," in which he argues, well, pretty much what the title describes.

Specifically: it's a common lament among academics that an increasing share of PhDs are unable to find full-time (i.e., "tenure-track") faculty positions and that an increasing share of courses (or credit hours) at major universities are being taught by temporary (or "contingent") faculty--who are often paid considerably less and have far less job security. An obvious question, then: if things are so bad for academics, then why do so many people continue to go to graduate school--particularly in the most desperate humanities fields--to get PhDs that are unlikely to yield full-time academic employment?

Afonso's creative and thoughtful answer is that it is the same motivation facing "rank-and-file" drug dealers: the prospect of success (i.e., being a gang boss or obtaining a tenure-track position) is worth the low pay and high insecurity that comes with being at the bottom. In other words, an ever-increasing supply of students are willing to go through the years of deferred income and stress of graduate school and the academic job market to be one of the ones who lands the coveted tenure-track job.

This is a smart argument, and well worth the read. In addition to this basic insight, Afonso further highlights how the different structures of academic in Britain, Germany, and the US further affect these dynamics. It has been discussed and reposted in dozens of forums. But let me add a different (and less exotic) perspective.

Nothing But an Adjunct, I Suppose
Despite the keen insights of Afonso's argument, there are some key differences between drug dealers and academics. Academics are quite unlikely to die or be arrested, and a tenure-track position bring a comfortable, but certainly not large, salary. By contrast, the allure for drug dealers is the prospect of a lot of money offset by a very high risk of arrest or death. Tenured academic employment brings the prospects of incredible job security with comfortable income. And therein lies the real problem for the academic job market.

Academic tenure is as close to absolute job security as one can find. There are ways to fire tenured faculty, of course, but it is very difficult and often litigious. In truth, university administrators have to make a careful decision when choosing to hire someone on a "tenure-track" line: a university can well be stuck with a tenured professor (and, moreover, the obligation to pay his/her salary and benefits) for decades. This fact, in turn, makes administrators very cautious to hire faculty on tenure-track lines. This is not just because of the (vastly overstated) potential for the "deadwood" tenured professor. It is also because things change: during the course of a 20-30 year academic career, changing events or student preferences may mean that there is little need or demand for the research or teaching expertise that the professor has. Nonetheless, the university still has to pay the professor, and it is difficult and costly to force the professor into early retirement. As a result, there are not many tenure-track faculty lines available in good years and almost none when times are bad.

This situation bears little in common with drug gangs, but it is quite similar to the shape of general labor markets in France and various other continental European states.

In France, employment is structured around the "indefinite term contract" (CDI), which heavily protects the employee from termination and mandates generous severance benefits in the case of termination (if not for cause). French employers thus face the same concerns as university administrators: hiring an employee on a CDI can turn out to be an expensive long-term proposition. Employers are reluctant do so, and the result is a chronically high unemployment rate--even in good times. (Another commonality is the effort to push early retirement on employees as a way of reducing labor costs or opening up new positions).

Moreover, successive French governments have responded to this problem in a similar way as American university administrators: by creating and expanding new contingent classes of employment that lack the job security and benefits of the CDI. The French job market ends up looking a bit like the academic job market: a group of secure "insiders" who are employed on the CDI, and a growing class of insecure "outsiders" who are employed on temporary or fixed-term contracts.

In short, the structure of employment law in academia and in countries like France contributes to this growing "dualization" or "insider-outsider" dynamic. High levels of job security generate higher overall levels of unemployment and lead to pressure from employers to create alternative employment contracts that bypass this job security. Academics who lament the decline of tenured faculty positions would do well to understand the nature of this problem.