Streetwise Professor

A recent RFS paper by Brogaard, Riggenberg, and Sovich purports to show that the rise of commodity index investing has had adverse effects on the real economy. Like most of the papers that analyze index investing, this one is seriously flawed and does not support the broad conclusions it asserts.

Like virtually all of the papers in this literature, it relies on crude before-after comparisons based on some magic year (2004? 2005? 2006?) during which commodity index investing became important. Even assuming that the rise in index investing was a sort of exogenous shock, this crude method of attributing causation has problems. After all, a lot of other stuff happened after, say, 2004. The rise of China as the predominate force in commodity markets, for instance.

Perhaps this problem would be less troublesome were commodities assigned randomly to “treatment” (part of an index) and “non-treatment” (non-index) groups. But this is definitely not the case. There are systematic differences between index and non-index commodities, so it is difficult to know whether the effects documented in Brogaard et al are due to a correlation between these systematic differences and the performance measures they utilize.

Ah. The performance measures. That raises yet more issues. The paper claims that firms with exposure to index commodities experienced lower returns on assets (i.e., operational income/assets) post-indexation than their non-index-exposed counterparts.

So what is the implicit model of the market in which these firms operate? If the presumption is that the markets are relatively competitive, why would you expect profit margins to change over a period of several years? Even assuming that indexation increased costs via the channels posited by the authors, in equilibrium this would lead to (a) an initial decline in profits, (b) exit (meaning fewer assets and lower output), and (c) a return of profits to the competitive level. Meaning that actually looking at assets or output would be a better measure of how indexation affects cost than profit margins.

If, conversely, the presumption is that these markets were imperfectly competitive prior to the indexation shock, the fall in profit rates could be a good thing, rather than a bad thing. It could indicate that indexation resulted in more intense competition/lower market power.

This is particularly interesting, given that this was a period in which measured profit rates were rising generally in the economy, a phenomenon which some (not I, but some) attribute to declining competition. Now, there are big problems with that literature (problems that the late great Harold Demsetz identified in the 1970s, but which modern scholars have forgotten), but take it at face value. Declining profits/margins could be interpreted as a signal of increased competitiveness and efficiency. A feature, not a bug.

And via a channel exactly contrary to that posited by Brogaard et al. They posit that indexation reduced the informational efficiency of commodity prices. But an increase in informational efficiency would tend to reduce market power–and reduce margins/profits. Note that many firms HATE the introduction of futures, or increased trading of futures, of their main inputs or outputs. Precisely because futures trading increases the informational content of prices which undermines an information asymmetry that generates profits for major producers and consumers.

Which brings me to that information channel. Brogaard et al measure informativeness using the autocorrelation in commodity returns.

Really?

They basically find that autocorrelations went up, which they interpret to mean that commodity futures markets became weak form inefficient (or further from weak form efficiency).

Again: really?

Taking their argument literally, it means that speculators were able to eliminate predictable movements in prices prior to indexation, but weren’t able to do so afterwards. This stretches credulity to its limits. After all, “financialization” of commodities also saw the entry of banks and hedge funds into commodity trading. We’re supposed to believe they left easy money on the table?

This purported reduction in price efficiency raises another issue. When I read about return autocorrelations I think time-varying expected returns. So the Brogaard et al result suggests that index commodity returns became more time-varying after indexation became a thing.

Arguably the primary impact of indexation was to integrate more closely commodity prices and stock and bond prices, and in particular, ensure that systematic risk was priced more consistently across commodities and financials. We know that equities and bonds have time varying returns. It is plausible that the documented increase in time variation in expected commodity returns reflects this integration, and is yet another feature rather than a bug.

If so, commodity prices that more accurately reflect the pricing of risk should lead to better investment decisions, not worse decisions. Further, these better investment decisions need not be associated, over several years, the operational performance measures they employ.

Indeed, along the lines of a paper that I wrote a few years back, indexation improves the allocation of risk and reduces the cost of commodity price risk to firms. In equilibrium, this would lead to lower rates of return. Which is exactly what Brogaard et al find.

Pursuing this further, I note that the authors do not use sensitivity to commodity prices as a means of determining whether firms are exposed to commodity prices, on the basis that hedging can reduce price exposure. Yes, but consider the following. Hedging is costly. Hedgers frequently pay a risk premium to speculators. This leads them to hedge less, which leads to high exposure. If indexation reduces hedging cost (as my paper implies), at the margin, firms will hedge more, and bear less risk. This reduces expected profits (which incorporate compensation for risk).

This has at least two implications. First, the average profit rate of firms exposed to hedgeable commodity prices (which are precisely the commodities that are included in indices) should decline post-indexation: this is consistent with the Brogaard et al finding. Second, firms should hedge more–leading to lower exposure to commodity prices. Brogaard et al do not test this latter implication. (Admittedly, this raises complications. Firms may limit commodity exposure in ways other than hedging, and a lower cost of hedging can lead to a substitution of hedging for these other means, leading to offsetting effects which reduce the commodity exposure impact of cheaper hedging.)

The upshot of all this is that the conclusions that the authors advance–namely, that indexation resulted in poorer real outcomes in terms of performance and investment–are not supported by their empirical findings. That is, there are alternative hypotheses that are consistent with the empirical findings that are diametrically opposed to their hypothesis.

This paper is well done within the analytical confines that its authors select. But that’s exactly the problem with this literature. The analytical confines exclude important channels of cause and effect. Most importantly, these confines make implicit assumptions about the degree and nature of competition that either make their results problematic, or mean that their results have diametrically opposed implications to those of the chosen analytical framework.

Meaning that the debate on the effects of indexation are still very much open, and that finance and economics scholars have largely failed to devise reliable tests to distinguish indexation-is-bad hypotheses from indexation-is-good hypotheses. Which is precisely why the debate is still open.