Back to Portfolio for the Future™

Back to the 1980s for future market insights

Christina Zhu, of the University of Pennsylvania, in a recent paper, looks at the consequences of Big Data for corporate management.

What “Big Data” means depends on who is talking or writing about it. For economists working within the classic microeconomic framework, it means a drastic fall in the cost of acquiring information. As Zhu observes, in recent years, taking advantage of the new technological landscape, a number of startups have gone into the business of collecting “alternative” data (clickstream info, satellite imagery of the cars in the parking lots of retailers, etc.) and presenting it to actual or potential investors.  Those investors now have available to them, though not without expense, a range of information they would never have dreamt of trying to compile as recently as a decade ago: granular and real-time data that does not depend on disclosures through the firm whose stocks or bonds they are considering.

Zhu shows that the consequences vindicate a model developed in the early 1980s, the “noisy rational expectations” (NRE) model. One of the key papers in the development of the model was that of Sanford J. Grossman and Joseph Stiglitz, “On the Impossibility of Informationally Efficient Markets.” Grossman and Stiglitz, writing in the days when clocks had hands and telephones were generally still attached to a wall, wrote that because information is costly, asset prices can never reflect all the information that is publicly available. This implies that there will always be room for players working to take advantage of the inefficiencies, for alpha seekers.

Douglas Diamond and Robert Verrecchia, writing in the same period, discussed how the asset price’s equilibrium serves as a “noisy” aggregation of the total information observed by all traders. There is a lot of static amidst the signals.

Times Have Changed

Zhu is saying: (a) they were right; and (b) times have changed. A decrease in the cost of information acquisition has allowed those new information intermediaries to spread these new types of data, which has in turn allowed investors, especially sophisticated investors of the sort most likely to make good use of such intermediaries, and, for example, to anticipate earnings announcements. Thus, as an empirical matter, Zhu says, “price reactions to earnings announcements are muted after alternative data from these data sources” have become available.

Of course, some firms get more thorough coverage from the alternative data intermediaries than others. This fact suggests Zhu’s method. She does a difference-in-differences study that compares 266 firms covered by alternative data sets with “a group of matched firms that are economically similar but do not have much data coverage.” For the firms with better coverage via the intermediaries, there is a measurable increase in the informativeness of prices. There is, to use again the terms of those early-Reagan-era studies, much less noise in the workings of rational expectations.

After satisfying herself on such points, Zhu then asks: what consequences does this have for managerial behavior? She has found evidence that: (1) Big Data discourages “opportunistic trading” by managers; and (2) it improves the company’s investment and divestment decision making.

Opportunistic Trading

To focus on the former: in an old-school early-1980s situation in which the numbers of an upcoming earnings report is a mystery to the rest of the world, but is known to the managers, those managers can trade on that information, extracting rents from their shareholders. The ability to extract rents would seem, a priori, to be inversely related to the information value of the pre-announcement stock price.  Zhu tests empirically whether “increased price informativeness disciplines the managers’ decision to trade and the directional magnitude of trades.”

Here answer is: yes. In quantitative terms, “the predicted probability of insider purchaser activity ahead of earnings increases is 17.7% lower for Covered companies after alternative data availability.”

The effect, though real and significant, is asymmetrical. In theory, insiders could extract rent either from selling or from buying ahead of an earnings announcement, depending, of course, on the direction of the surprise (if any) the announcement was to contain. But the effect of alternative data sets on insider sales ahead of earnings decreases is insignificant. She attributes the asymmetry to the fact that insiders “face more constraints and litigation risk” on sales than on buys. Hence, they already are deterred from making such a sale, and Big Data has less work to do here.