Friday, March 7, 2014


There is a very interesting paper that recently got published in the Econometrica using a Regression Discontinuity design. The focus is the impact of Islamic rule on the education choices of young women.
Meyersson uses Turkish elections data and focuses on a particular time in history, 1994 Turkish elections, that brought Islamists to power in the mayoral elections. In economics it is difficult to separate causality from correlation. Say, if there are fewer women going to school in a district that an Islamist party governs, is this because of the Islamist party rule or did the Islamists come to power simply because the region is dominated by an Islamist minded populace where it is the norm for girls to drop out of school after puberty?
Meyersson first shows the expected, i.e. that there is a negative correlation between an Islamist win in a region and the school completion by women. Again, remember this is not causality but rather correlation. Then he classifies mayoral districts by the percentage win/loss of Islamists and focuses on the discontinuity of an Islamist win vs an Islamist loss, say comparing districts where Islamists lost the election by a small margin of 1-2% vs. the districts where Islamists won by a margin of 1-2%. These are the districts where you can take a better snapshot of the real Islamic influence on girls' education, as the district could have gone either way. If it did go Islamic then how did this influence girls' education vs. if it did not? That's an excellent question and great research design. As you can imagine Islamists shouldn't have a positive or negative influence in the municipalities where they lose the elections. (this might actually be more complicated, but I will try to be brief in this post)
Meyersson finds that in fact girls' school attendance beyond puberty as well as middle school and high school completion rates increase in regions that were narrowly won by Islamist parties. This is a very striking result. In Turkey municipalities do not build or administer schools. Meyersson offers two potential causal explanations of why girls' school attendance increased under Refah:
1. That, with a new local government, the state schools became more open to allowing girls with headscarves to attend classes against the central government's wishes.
2. With the Refah municipality in power private religious schools and vakif schools became more active in these districts.
It's an excellent paper.

Friday, February 28, 2014

Research Ideas - 3 /

Firm Complexity and Post-Earnings Announcement Drift

This paper has been written with Alexander Barinov and Shawn Saeyeul Park. The paper shows that the post-earnings-announcement drift (PEAD) is stronger for conglomerates, despite conglomerates being larger, more liquid, and more actively researched by investors. We attribute this finding to slower information processing about complex firms and show that the post-earnings-announcement drift is positively related to measures of conglomerate complexity. We also find that the post-earnings-announcement drift is stronger for new conglomerates than it is for existing conglomerates and that investors are most confused about complicated firms that expand from within rather than firms that diversify into new business segments via mergers and acquisitions.

In this paper we build on the "Complicated Firms" paper written by Cohen and Lou (2012). Cohen and Lou (2012) show that industry-wide news take a lot longer to get incorporated into the prices of conglomerates than for single-segment firms. In this paper we show that earnings-related information is another type of information investors in conglomerates have trouble digesting. 

While doing so we reach three main conclusions: First, the stronger PEAD for complex firms is independent from the return predictability documented by Cohen and Lou (2012) and thus represents a separate case of the impact of firm complexity on stock prices. Second, roughly a quarter of the relation between PEAD and complexity can be attributed to the fact that a unit of SUE has more information for complex firms than for simple firms and another quarter of the relation can be attributed to the relatively low analyst coverage of complex firms after one controls for size. Third, even after controlling for these alternative explanations,complexity per se plays an important role as a limits to arbitrage variable.

While writing this paper perhaps what I was the most surprised about was discovering that single-segment firms receive a lot more analyst coverage than multi-segment firms once one controls for size. This is true whether one uses the number of analysts or the percentage of industry experts as his/her measure of analyst coverage. We plan to expand on this finding further in future papers.

Research Ideas -2 / Should you rely on Major Customers?

Should you rely on Major Customers?

Customer-base concentration, profitability and distress across the corporate life cycle

This paper has been written with Paul Irvine and Shawn Saeyeul Park. The paper focuses on better understanding an age old question in economics and finance. Is it better to depend on a few, large, well-established customers or should you try to expand your customer base? Both strategies have advantages and disadvantages. Dealing with a few large customers could lead to more efficient supply chain management and better visibility into future orders. On the other hand an asymmetrically powerful major customer could demand price concessions, leading to 

​significantly smaller gross margins.  Whether the advantages or disadvantages of relying on a few major customers dominate is entirely an empirical question. Our analysis confirms several well-established facts: We verify that relying on a few, major customers can hurt you as such a reliance leads to lower gross-margins. We also verify that there are benefits to relying on a few, major customers as suppliers with high customer-base concentration have lower inventory costs and better cash and inventory turnover. While these facts have already been well known in the literature we discover two entirely new facts about suppliers that rely on a few, major customers which improve our understanding of how supply chain relationships affect firm profitability and survival. 

First, we find that supplier firms that rely on a few, major customers make much larger customer specific investments evidenced by the much higher rigidity of their costs. Such a rigid cost basis implies that when things are good with their customers (i.e. when their customers are doing well) supplier firms with concentrated customer bases will do even better as they need to spend a lot less for generating an additional unit of revenue. However when times are bad, sales to major customers will shrink and since suppliers with concentrated customer bases are stuck with the fixed investments incurred earlier, in bad times a supplier with a concentrated customer base will suffer considerably more than a supplier firm with a diverse customer base. This is so because suppliers with diversified customer bases have a lot less in fixed investments and as such they can reduce their costs much more efficiently in bad times. This operating leverage effect cuts both ways. 

Second, we find that firms that rely on a few, major customers have a lot more demand uncertainty, not to be confused with demand visibility. In portfolio theory we teach our students that holding a larger number of (hopefully not perfectly correlated) assets reduces the overall volatility of their portfolio. The same idea applies to sales volatility. Relying on a few, major customers may help with demand visibility as the supplier firm could be better informed about upcoming orders. This certainly would reduce inventory holdings. However, such visibility wouldn't decrease the volatility of sales for such a supplier, as shocks affecting the major customers would directly be transmitted to the supplier firm: Relying on a few major customers leaves the supplier exposed, without the possibility of making up for lost sales through alternative channels and leads to significantly larger demand uncertainty (sales volatility).

Thus we find that suppliers with concentrated customer bases are firms with large operating leverage and high demand uncertainty. Such firms have a lot more likelihood of facing "bad" states of the world with no ability to reduce costs in such states and thus are a lot more likely to fail. Should they survive the "bad" states of the world suppliers with concentrated customer bases can enjoy certain operational benefits such as reduced inventory holdings, higher cash turnover as well as higher inventory turnover. We conclude that customer concentration brings both costs and benefits to supplier firms. ​

Research Ideas - 1 / Pricing of Default Risk in Equity Markets

Is there a distress risk anomaly? Pricing of default risk in the cross-section of equity returns

This paper has been written with Deniz Angıner. It’s about the so-called default risk premium anomaly. The basic idea is that if default risk is a systematic risk then investors should get compensated for being exposed to this risk. The fundamental thinking is not any different from when we teach about the CAPM (Capital Asset Pricing Model). Then why bother writing this paper and what do I mean by the default risk premium anomaly? Well it turns out that using traditional measures of default risk resulted in unexpected findings, i.e. firms with higher default risk seem to under-perform firms with lower default risk in the next period. But then why would investors expose themselves to this type of risk if they do not get any compensation in return?

We are the first paper in this literature to point out that traditional measures of default risk are not necessarily good measures of exposure to the systematic component of default risk. Think about CAPM beta vs. the firm’s return volatility: For which of these measures do you expect to get compensation? Theoretically you would expect a premium due to your co-variation with the market , i.e. you’d get a premium based on your CAPM-beta not based on your return volatility, i.e. a firm can have very volatile returns but these returns may co-vary very little with the market returns. Similarly a firm may have a high expected default probability but this default probability may not be correlated with the default likelihoods of other firms in the market. And if that is the case (a high default probability with little correlation with other defaults in the market) then the investor wouldn’t necessarily be compensated for this firm-specific risk. In this paper we devise a method to better understand a firm’s covariation with the common default risk factor rather than blindly using a firm’s real world probability of default. Anginer and Yildizhan (2013) has been around for a while now, we have been downloaded over 1,000 times in SSRN, Our paper has already been cited more than 17 times in articles and books, and we hope for it to be published in a good journal soon.