Responses to AFL-CIO’s Critique of the Agrawal Study

Ashwini Agrawal wrote to us:

In my study, Corporate Governance Objectives of Labor Union Shareholders, I examine the proxy votes of the AFL-CIO Reserve Fund and Staff Retirement Fund. I compare the votes before and after the AFL-CIO split into two groups: the AFL-CIO and the Change to Win Coalition. After the breakup, the funds become relatively more supportive of directors of firms in which workers become primarily affiliated with the Change to Win Coalition.

In his blog post and letter, Daniel Pedrotty makes a number of mischaracterizations regarding my study.

For example, in the paper I note that the AFL-CIO proxy votes are cast by a third party fiduciary in accordance with AFL-CIO proxy voting guidelines; I do not argue otherwise, as Pedrotty claims. AFL-CIO proxy voting guidelines can be found here.

In another example, Pedrotty states that changes in AFL-CIO voting patterns solely reflect changes in governance characteristics. To investigate this, I look at the votes of other institutional investors who take governance characteristics into account, such as mutual funds Fidelity, Vanguard, TIAA-CREF and union pension funds associated with the Brotherhood of Carpenters (UBCJA). I find that these other investors do not change their voting patterns in the same ways as AFL-CIO funds, suggesting the AFL-CIO votes are not solely based on governance characteristics.

Pedrotty also erroneously claims I treat all firms with mixed union representation as though they are affiliated with the AFL-CIO only. Based on publicly available data, I distinctly categorize a company based on whether a significant fraction (at least 90%) of the unionized workers at the firm switches affiliation from the AFL-CIO to the Change to Win Coalition. This is meant to capture significant decreases in the AFL-CIO’s labor representation across firms. For example, if 50% of a firm’s unionized workers remain part of the AFL-CIO, while the other 50% switch to the Change to Win coalition, then I assume the AFL-CIO still has substantial labor representation in this firm. If labor interests do not influence proxy votes, this categorization should not impact the findings in the paper.

Pedrotty asserts I gather data on certain Change to Win funds but that I do not compare their voting behavior with the AFL-CIO. This is incorrect; I compare the UBCJA pension fund votes (a Change to Win affiliate) with those of the AFL-CIO funds. Upon joining the Change to Win Coalition, the UBCJA funds become relatively more opposed to directors of firms in which workers are primarily affiliated with the Change to Win Coalition.

Pedrotty also claims that I made no effort to inquire into the methods by which AFL-CIO proxies are voted. When I requested this information from the AFL-CIO Office of Investment (in May, 2007), I was told by the AFL-CIO Office of Investment that this information would not be disclosed to me (in June, 2007).

The AFL-CIO’s claim that their voting fiduciary cast votes the same way for certain CTW funds does not change the findings as they pertain to AFL-CIO affiliated funds. In addition to the AFL-CIO not disclosing the sizes of these CTW funds and the extent to which AFL-CIO proxy voting guidelines are applied to them, the AFL-CIO’s claim could reflect agency issues within the voting structure of union pension funds. However, making any such assessment is beyond the scope of my paper.

This study can be easily replicated using publicly available sources of data. Both the sources and methodologies are described in the paper, available online here. It should also be noted that I use publicly available data because it is verifiable by other researchers and because my requests to the AFL-CIO for access to their own database on union membership were denied. I am happy to incorporate the AFL-CIO’s internal data into my study, however this data has still not been given to me.

I encourage blog readers to read both my study and the AFL-CIO’s report to reach their own conclusions. Peer review, comments, and criticisms are welcome.

Steven Kaplan wrote to us:

I would like to respond to Daniel Pedrotty’s post on Ashwini Agrawal’s work. I am a member (but not the chairman) of Mr. Agrawal’s dissertation committee so I know his work well. His committee consists of four professors at the University of Chicago Graduate School of Business.

Daniel Pedrotty’s post (criticizing Ashwini Agrawal’s study) does not in any way contradict Mr. Agrawal’s study. Mr. Pedrotty claims the paper makes a “serious and completely false accusation of voting behavior.” That is not true. Mr. Agrawal’s paper reports the results of his data collection and analysis. Mr. Agrawal’s results use publicly available data and are replicable. Using data on AFL-CIO votes and data from public sources regarding AFL-CIO union representation, Mr. Agrawal finds significant changes in AFL-CIO votes. After the split of the AFL-CIO and CtW, the AFL-CIO pension fund was much more likely to vote for management and directors of companies with unions primarily represented by the CtW than before the split. At the same time, the AFL-CIO pension fund remained less likely to vote for management and directors of companies still mainly represented by the AFL-CIO. The union voting behavior that Mr. Agrawal’s analysis reports is more consistent with union self-interest than with shareholder value maximization. While any such behavior cannot be proved beyond a shadow of a doubt, the results are very statistically significant.

Mr. Pedrotty has not attempted to replicate Mr. Agrawal’s methodology or tests. He presents some other data and claims that are irrelevant to Mr. Agrawal’s analysis. And, therefore, Mr. Pedrotty’s post does not have anything substantive to say about Mr. Agrawal’s results. Given this, it is extraordinary that Mr. Pedrotty request Mr. Agrawal to withdraw or revise his paper.

Let me examine Mr. Agrawal’s paper and Mr. Pedrotty’s post in more detail.

First, let’s be clear about what Mr. Agrawal did. He asked a very interesting question. Is union voting consistent with shareholder value maximization or union self-interest? He came up with a very clever way to test this. He focuses on the split-off of the CtW (Change to Win) from the AFL-CIO in 2005. He considers whether the voting patterns of the AFL-CIO pension fund change depending on which union – the AFL-CIO or the CtW – is the primary worker representative at a particular company. Note, that the question is an empirical one. The paper would be very interesting regardless of what result he found. That is why the committee thought the project so promising.

Mr. Agrawal collected the data on shareholder votes directly from the AFL-CIO and from one CtW affiliated fund. I believe those votes are now publicly available. Mr. Agrawal consulted 10-Ks and annual reports, FMCS F-7 notices, etc. to determine which unions represented employees at which companies during the sample period. He relied exclusively on this publicly available data. Mr. Agrawal details how he measured union representation and how he measured voting behavior as votes ‘for’ or ‘against’ directors. As a result, his study is easily replicable by other researchers. This is an important point. It does not rely on data that can be shaded by an interested party.

Mr. Pedrotty has chosen not to replicate the study and has asked for Mr. Agrawal’s data. It would be highly unusual for Mr. Agrawal to turn over his data set at this point. It is not customary for researchers in finance and economics to release their data to other researchers (particularly potentially hostile ones) before a paper has been officially published. I have rejected many such requests for data. And I rarely make such requests of others. In putting together a data set, a researcher spends a great deal of time and effort. It would be inappropriate to just give that data away for free to other researchers before fully utilizing and benefiting from the data collection effort. It also is worth pointing out that the AFL-CIO pension fund refused to give Mr. Agrawal certain data he asked for last year.

Using data on AFL-CIO votes and data from public sources regarding AFL-CIO union representation, Mr. Agrawal finds significant changes in AFL-CIO votes. After the split of the AFL-CIO and CtW, the AFL-CIO pension fund was much more likely to vote for management and directors of companies with unions represented by the CtW than before the split. Votes shifted from roughly 50% against management to only 30%. At the same time, the AFL-CIO pension fund remained less likely to vote for management and directors of companies still represented by the AFL-CIO. Votes remained at 50% before and after. As a control, Mr. Agrawal finds that other mutual fund investors exhibited no such change in behavior. The one CtW fund for whom he could get separate voting info voted in the opposite manner. I.e., the Brotherhood of Carpenters (a CtW affiliate) became more likely to vote against directors of firms represented primarily by the CtW.

These patterns are difficult to reconcile with shareholder value maximization. They are much more consistent with union self-interest. And to repeat, Mr. Pedrotty’s post in no way challenges Mr. Agrawal’s results.

Now, let’s look at the points Mr. Pedrotty does choose to make.

First, Mr. Pedrotty claims that “Agrawal has refused to discuss his findings, disclose his data set, or respond to substantive criticisms of his paper.” I am confused by this. Mr. Agrawal has presented his paper in academic seminars across the country – including Duke, Harvard, New York, and Stanford Universities. As discussed above, Mr. Agrawal relied on publicly available data or data (on shareholder votes) received directly from the AFL-CIO. It is clear what Mr. Agrawal did. And his results can be replicated. Pedrotty did not try to replicate Mr. Agrawal’s results. As mentioned above, it would be highly unusual for Mr. Agrawal to turn over his data set at this point.

Second, Mr. Pedrotty claims the paper makes a “serious and completely false accusation of voting behavior.” That is not true. The paper reports the results of Mr. Agrawal’s data collection and analysis. The results are more consistent with voting behavior in the AFL-CIO’s self interest than it is with shareholder value maximization. While any such behavior cannot be proved beyond a shadow of a doubt, the results are very statistically significant. It is perhaps useful to mention that option backdating also was suggested by a serious and replicable statistical analysis (although those analyses did not prove such behavior, and not all firms identified as backdating engaged in that behavior).

Third, Mr. Pedrotty asks for a peer review by experts. That will happen naturally when Mr. Agrawal submits the paper to an academic journal. He will make such a submission after considering all the comments he receives, including those of Mr. Pedrotty. In the meantime, it is worth pointing out that his dissertation committee is chaired by Marianne Bertrand – winner of the Bennet Prize in 2004 for the top woman economist in the world under 40 for work in labor economics and corporate governance – and includes Josh Rauh – winner of the best paper award in corporate finance in the Journal of Finance in 2006. I am curious as to what other types of experts Mr. Pedrotty has in mind.

Fourth, Mr. Pedrotty claims Mr. Agrawal uses voting behavior of mutual funds to define governance characteristics rather than voting recommendations of more independent sources. I must say I do not understand this point. The key point from Mr. Agrawal’s research design is that he looks at votes for and against management and directors. Mr. Agrawal is clear on this point. Mr. Agrawal also clearly finds that the AFL-CIO voting behavior before and after the CtW split is very different. Nothing Mr. Pedrotty says changes these facts.

Fifth, Mr. Pedrotty claims that firms with mixed union representation are treated as AFLCIO firms only. That is false. Mr. Agrawal uses publicly available data, and categorizes a company based on whether at least 90% of the unionized workers at the firm switch affiliation from the AFL-CIO to the CtW Coalition. This will capture decreases in the AFL-CIO’s labor representation across firms. Thus, Mr. Agrawal relies on a sensible, replicable, and factual methodology.

Sixth, Mr. Pedrotty claims that the AFL-CIO and CtW affiliated funds vote in the same manner. That is just not true in the case of the one CtW affiliated fund that provided data to Mr. Agrawal. With regard to other funds, that is possible and Mr. Agrawal may mention that in the next draft of his paper. Nevertheless, that does not change the fact pattern of the vote change for the AFL-CIO funds.

Seventh, Mr. Pedrotty claims that Mr. Agrawal did not compare the voting behavior of AFL-CIO and CtW affiliated funds. That is simply wrong. That is what the paper does in a statistically sophisticated fashion.

In sum, many of Mr. Pedrotty’s claims about Mr. Agrawal’s paper are false. Others are irrelevant to Mr. Agrawal’s paper. Mr. Pedrotty has in no way contradicted Agrawal’s results. In fact, Mr. Pedrotty does not even address them. In other words, Mr. Pedrotty’s post does not have anything substantive to say about Mr. Agrawal’s results. Given this, it is extraordinary that an interested party such as Mr. Pedrotty ask a careful and independent researcher such as Mr. Agrawal to withdraw the paper.