Thursday, May 04, 2006
U.S. companies donate significant amounts of money to politically affiliated organizations every year. Assuming that, over the long run, there is a similar likelihood that a Democratic or Republican will win the presidency (or at least that there is a significant amount of uncertainty), it would seem that, on average, profit-maximizing firms would wish to hedge their bets and give roughly equal amounts to both parties. While this is, no doubt, a common approach, firms choose to give disproportionately to one party with surprising frequency. While there are many possible reasons they might wish to do so, this paper seeks to test one hypothesis, namely, that CEOs use company money to support personal, political interests over and above the interests of the firm and its shareholders. If able to get away with it, using firm money to support political interests could make sense from a self-interested CEO's perspective since the benefit derived from giving firm money is likely similar to that of giving personal money, while the costs to the CEO are significantly lower.
Prediction: If CEOs use company money to support personal political interests, the political giving of their firm should be highly correlated with the political affiliations of the firm’s CEO. (See below for concerns)
Data & Empirical Strategy:
This paper uses two separate datasets. The first is based on information from “Who’s Who?”, a biographical database of world “leaders and achievers,” and includes information on over 2,700 firm executives and, in some cases, their political affiliations. The second contains information on various types of political donations by thousands of U.S. firms.
Looking to estimate the effect of CEOs political affiliation on political donations by firms…
Proportion of overall donations to Democrats= a + B(ExecDemocrat) + E
Where ExecDemocrat is a binary variable =1 if CEO is Democrat, 0 if not
Proportion of overall donations to Republicans= a + B(ExecRepub) + E
Where ExecRepub is a binary variable =1 if CEO Republican, 0 if not
Initial Concerns & Issues:
The most serious concerns relate to OVB/endogeneity. Specifically, if a firm is largely Republican, it might be more likely to hire a Republican CEO. In such a case, the fact that the firm gives disproportionately to the Republican party is obviously not solely attributable to the CEO, which would result in our estimate overstating the true effect. Additionally, it is likely that there are a number of other effects affecting proportion of donations given to a certain party and correlated with our Uis, resulting in OVB. Ideally, we would like to control for firm size, industry (maybe oil companies or large companies give to Republicans regardless of CEO’s affiliation), and, if possible, firm political affiliation.
Sampling error is also a concern. For the data on 2,760 executives, political affiliation is “not specified” for 80% of executives and “missing” for 9%. If there are reasons some chose not to specify (i.e., their being specified or not was not random), our sample of Democrats and Republicans might not be representative of the population of interest, leading our estimates to be biased.
Good summary. I like the topic, but, as you note, you have some hurdles to overcome.
1) Data -- I am surprised that you have the political affiliations of so few CEOs. Isn't the party that you include on your voter registration a matter of public record? I would have thought that some watchdog type group would have kept track of this stuff. If you want, you could probably (imprecisely) back this out of the FEC data. We could take your list of CEOs and match it to the individual giving records from 1980-2004. Then you could try and infer party affiliation of the individual from their giving prior to becoming CEO. The problem with this approach is that one of the ways that companies get around limits on contributions is by having their employees contribute. So some of these people's contributions may not reflect their own views.
2) ID Strategy -- one way improve the credibility of your estimates would be to use firm fixed effects. Do you have, or can you put together, a panel dataset of firm giving and CEO affiliation? If so, using firm fixed effects would allow the estimation of how giving patterns change when the party affiliation of the CEO changes. While not solving all the potential problems, this, at least, reduces some of the OVB concerns.
Subscribe to Post Comments [Atom]
Subscribe to Posts [Atom]