(This post was originally featured in the Review of Finance Managing Editor’s blog, which summarizes the lead article of each issue in a non-technical manner.)
John Cochrane of Stanford has written an excellent review of recent advances and future research directions in macro-finance. Macro finance studies the relationships between asset prices (e.g. the level of the stock market) and economic conditions (e.g. whether we’re in a recession or a boom). These relationships are important. For example, if investors are more willing to bear risk in good times, stocks will be more attractive, driving up the price and leading to lower future returns. Thus, current asset prices, and future expected returns, will vary with economic conditions. Conversely, current asset prices, and past asset returns, can forecast future economic conditions such as GDP growth and inflation.
Why might investors’ willingness to bear risk depend on economic conditions? To answer this question, let’s go back to why risk matters to begin with. To be willing to hold stocks, an investor must be offered a high expected return – not because stocks are risky (their value may go up or down), but their value may go down at particularly inconvenient times – times in which the investor is in real need of money (more technically, the marginal utility of money is high) and so is particularly hit by the stocks’ poor performance. For example, a recession is a bad time – the investor may have lost his job, and suffers a double whammy if his stocks do badly also.
The goal of macro-finance is to identify what these “bad times” are. Doing so allows us to answer many other big picture questions. If investors’ willingness to bear risk varies with these identified “good/bad times”, so will asset prices and expected future returns. This may in turn explain why:
- Asset prices are so volatile, as documented by the line of research that ultimately won Bob Shiller the (joint) 2013 Nobel Prize in Economics.
- The equity premium is so high – why stocks offer so high returns compared to Treasury bills – because stocks may perform badly precisely at bad times.
- Asset returns are predictable, i.e. “good/bad times” predict future returns.
The standard consumption-based asset pricing model argues that good/bad times are defined purely by consumption growth. Bad times are times in which consumption is low, and so the investor is in real need of money. If stocks do badly in times when consumption is low, then stocks are risky and investors will demand a high expected return. But, consumption just isn’t volatile enough in reality to explain the high equity premium we observe in the data – the famous equity premium puzzle of Mehra and Prescott (1985).
Macro-finance models thus identify forces – other than consumption growth – that may affect marginal utility. The paper studies several different such forces. I summarize a subset of the key ones here, but the paper goes into much more detail on the evidence for and against each set of models, and in particular the future research that can be done to further support or rule out each explanation.
A “habit” is a minimum level of consumption such that, if the investor’s consumption falls below that level, he suffers great disutility (unhappiness). In an “external” habit model, this consumption may be the consumption of other people, as in “keeping up with the Joneses”. In an “internal” habit model, this consumption may be your own past consumption – a student may be happy living off cold pizza, but once she becomes an investment banker and is used to sushi, she can’t fathom going back to cold pizza. External habit models – a seminal paper being Campbell and Cochrane (1999) – are particularly tractable (easy to model) because the investor’s current consumption doesn’t affect his future habit.
These habits mean that an investor’s risk aversion is time-varying. In bad times, his consumption falls precariously close to his habit level, and so he becomes extremely risk-averse. Thus, he is particularly unwilling to hold stocks, and so stock prices are low. Despite stocks appearing a “bargain”, they actually aren’t, since the investor is afraid of buying stocks, them falling, and him falling below his minimum desired consumption level. In good times, investors are more willing to hold stocks, and “reach for yield” as is observed in practice.
Importantly, habit models can explain not only a high equity premium, but also a low interest rate. In the standard consumption-based asset pricing model, where risk aversion is constant over time, you could generate a high equity premium (despite consumption not being that volatile) by arguing that investors are simply extremely risk averse. But (in addition to being counterfactual given observed investor behavior), this leads to a separate problem. If investors are very risk averse, they want to even out consumption over different years. The student really can’t stand eating cold pizza; knowing that she will become an investment banker in the future, she will borrow from her future income so that she can at least enjoy fresh pasta. But, such borrowing would drive the interest rate much higher than observed in practice. In habit models, the interest rate stays low because, even though the above borrowing motive exists, it is offset by precautionary savings motives – the investor wants to save to reduce the risk that his stocks do badly and his consumption falls below his habit level.
In long-run risk models – a seminal paper being Bansal and Yaron (2004) – “bad times” are not times in which consumption is low compared to the habit level, but times in which the investor also receives bad news about long-run future consumption growth. In other words, if stocks fall at times in which technology suddenly declines – meaning a permanent loss of productivity and thus future dividends on stocks – then the investor is again hit with a double whammy.
Conceptually, it is difficult to really believe that people were unhappy that their stocks fell in, say, Fall 2008, not because of anything having to do with the current economic disaster, but only because there was some news about far off future living standards. And the higher risk premium in Fall 2008 came only because of higher volatility about such long-run news. But verifying such news and the plausibility of the mechanism remains a hot research agenda in this line of work.
In idiosyncratic risk models – a seminal paper being Constantinides and Duffie (1996) – “bad times” are times in which the investor faces a lot of idiosyncratic (personal risk) to his consumption. Assume the investor is a professor coming up for tenure, and assume that the tenure process is random – which of course it is not, because in real life it is entirely based on merit and not at all on politics. Getting tenure is great, getting denied tenure is lousy, but because investors are risk averse, the pleasure of a great outcome (tenure) is less than the pain of a bad outcome (no tenure). Thus, times of big idiosyncratic risk are bad times, and if stocks just happened to go down when you come up for tenure, then stocks would be risky. But, this would require stocks to coincidentally go down precisely when lots of professors are coming up for tenure and so cross-sectional consumption volatility is high.
The mechanism is plausible – cross-sectional risks are higher in bad times. So far, the data do not seem to show enough variation in cross-sectional consumption volatility to explain the level and variation of the equity premium. But new work emphasizing the individual rare disasters shows promise.
In heterogeneous preference models – a seminal paper being Garleanu and Panageas (2015) – “bad times” are times in which investors who hold most stock become more risk averse. In idiosyncratic risk models, investors have the same preferences but face idiosyncratic risk. In heterogeneous preference models, investors don’t face idiosyncratic risk but have different preferences – specifically, different levels of risk aversion. Risk-tolerant investors hold more stock than risk-averse ones. But, when the market falls, the large stockholders lose more money. Thus, risk-tolerant investors comprise a smaller part of the overall market, and so the market as a whole becomes more risk averse – precisely when the market falls.
Intermediary Asset Pricing
In intermediary asset pricing models – a seminal paper being He and Krishnamurthy (2013) – intermediaries (such as banks) are the key investor, and “bad times” are times in which intermediaries are close to bankruptcy and so become risk averse. These models work much like habit models, where the habit (minimum level of consumption) is replaced by the level of debt the investor must repay. This requires unlevered investors (such as wealthy individuals, or university endowments) to be unable to buy stocks at the time they are cheap because levered investors are unable to.
In behavioral models, investors form expectations in irrational ways, and “bad times” are times in which expected future returns are (irrationally) low. For example, in Barberis et al. (2015), investors over-extrapolate from past returns. After the stock market has fallen, investors irrationally think that it will continue to fall (even if, in reality, past returns have no link to future returns). Thus, even though stock prices are low, they are not a “bargain”, because investors forecast expected returns to also be low.
Despite the necessary comparisons of the various approaches, really the theme of the paper is their underlying unity. They all describe state variables beyond consumption – reasons for high marginal utility in bad times – that drive the equity premium, and they all describe mechanisms for higher risk aversion in bad times. They offer much the same descriptions of aggregate data – consumption and stock prices. They are essentially different microeconomic mechanisms for the same macroeconomic phenomenon. They all inform macroeconomics that recessions deeply involve the size and variation of risk premiums. Which micro-foundation, or which combination of micro-foundations, wins out in the end will be the fascinating area for the next round of research, but it will build on this essential unity.
The field of macro-finance is far richer than the above brief summary can do justice to, and arguably the most interesting part of the paper (which occupies over a third of its length) is a “Research Agenda” highlighting new facts to be explained by future research, and other open questions. Since these open questions have not yet even been studied by experts in macro-finance, non-experts like me cannot possibly do justice to them in a summary, so I will refer the reader to the paper to these. But, I hope that this summary has piqued the interest of those outside the literature and encouraged some of you to read the paper.