Friday, June 27, 2014
With the SEC still sorta thinking about requiring U.S. public firms to disclose their political expenditures, a newly posted paper by Saumya Prabhat and David Primo (of the Indian School of Business in Hyderabad and University of Rochester, respectively) should be getting some play. Prabhat and Primo claim to find evidence that the U.K.’s enactment of similar disclosure rules in 2000 was bad for shareholders of politically active firms, and they argue this is a reason to oppose possible U.S. reforms. I think they’ve got at least one really big methodological problem, and even taking their results at face value they’re just misinterpreting them.
Let’s start with their findings. Prabhat & Primo report that news of the expected reforms increased stock-price volatility for politically-active firms (16) and that a portfolio of active firms performed worse after news broke than a portfolio of other firms (18). For some reason they emphasize the first much more, even though it’s kind of a no-brainer that there will be stock price volatility for regulated firms when the market gets information about a new regulation affecting them.
I’ll focus on the second finding. First, causation or correlation? As P&P themselves note, many firms lobby as a fallback when things go badly, or to close off entry by threatening new competitors. Others lobby to head off government investigations or regulatory actions. Is it surprising, then, that the firms that were lobbying just before the period P&P observe performed worse than other firms? Probably not: lobbying is itself a red flag that there’s a rocky road ahead. Or, at a minimum, it is a signal to the market that managers know there is a potential pothole looming.
Since this is a post about the value (or not) of disclosure, let me say that again. News about firm lobbying could provide information to the market about managers’ private knowledge of looming challenges for the firm. And nothing that P&P report is inconsistent with that story.
Here’s another way in which their results show that disclosure was good for shareholders. Let’s say that most lobbying by individual firms is firm-specific rent-seeking. (Firms may also sometimes be willing lobby for collective goods, but often that is done through trade associations). From a diversified shareholder perspective, all firm-specific rent-seeking is destructive. The lobbying is moving money from Firm A to Firm B. I’m invested in both. So the two firms are wasting resources to fight over which of my pockets my money is in. (Contrast this with money spent on, say, R&D, which does just shift pockets to some extent but also contributes to growth). And maybe I’m also a consumer. They’re trying to shut out new competitors, keeping the prices I pay high.
Disclosures allow me, the diversified investor, to identify which firms are wasting my investment in this way. You can understand why undiversified managers would prefer to enrich themselves at the expense of competitors, but this is exactly the kind of self-serving behavior that investors want to curtail. So, long story short, if disclosure is actually reducing firm value, that could simply represent the fact that shareholders are unwilling to invest in managerial self-dealing.
Prabhat & Primo half-heartedly try to reject these kinds of stories by claiming that they have tested for the effects of poor governance. They create a dummy variable for whether a firm has any large blockholders, and interact that with the enactment of disclosure rules (17). They find no significant effects, and proclaim that this means poor governance (and hence managerial self-dealing?) is not driving the results. This is a bit like looking in the hall closet, finding it dry, and declaring that it is not raining. There are many many more sophisticated ways of measuring governance quality (none perfect). Consider that governance may even be negatively related to blockholding--for example, if the presence of blockholders increases freeriding by other monitors, or if firms without blockholders take on other governance commitments to appeal to small investors.
And keep in mind we’re just talking about shareholder welfare here. As P&P acknowledge in their footnote 7, they aren’t conducting an overall social welfare analysis. But diversified shareholders probably care about that, too!
Finally, that methodology problem. PMost of the regression results are based on a standard difference-in-differences (“DD”) setup in which they examine the effects of the law before and after an event, comparing the treated group to a group of controls. The before and after they study is either the date of the report, the legislation two years later, or the whole period encapsulating both. Here’s the thing. Their treatment variable is firms that were politically active before the legislation was enacted. Note that I did not just say before the news of the impending regulation broke. I checked this; they say it twice (14, 16).
That isn’t how DD works. You can’t have a treatment drug and a placebo drug and let the study participants decide which one to take. What P&P are measuring is the effect of the new regulation on firms that decide to lobby even after they know it will be disclosed. (Some of the firms in their sample might have lobbied only before the first event, but they don’t tell us the mix.) So you have a big selection bias concern. For example, maybe the firms that lobby in the face of disclosure are the ones that are especially desperate to ward off whatever bad outcome they expect is upcoming, which would bias their results way upwards. Or, heck, maybe they’re selecting for firms that know they won’t be hurt by disclosure, in which case results are biased downwards. Either way, I don’t believe the results.
One last, technical nitpick about their DD approach. A necessary assumption for this method is that the treatment and control groups would have behaved similarly if not for the treatment. To show that, usually one presents a graph in which they have followed parallel trends for a long time before the treatment event. P&P instead show us a year and half of pre-event data, and even there the trends for firm-specific risk (27 fig. 2) don’t look all that parallel to me. This is just basic U.K. stock value data; it’s available back to like the 19th Century. Showing just 18 months raises my eyebrows.