At the heart of the intellectual revolution in macrofinance, which began in the wake of the global financial crisis, is risk appetite. The risk appetite of market-based intermediaries is the tidal force that drives systematic fluctuations in asset prices. It is the main state variable in intermediary asset pricing, meaning that it is controls whether the state of the world is good or bad. In that sense, it’s like the weather. It affects everyone interested in securities.
The Fed really cares about risk appetite because it is the principal channel through which monetary policy affects economic activity. Monetary tightening, say, through rapid rate hikes, for instance, drives risk appetite lower; resulting in tighter financial conditions, in turn reducing risk-taking and the pace of credit creation. This then cools the economy, tempering inflation. The central role played by risk appetite was poorly understood before the global financial crisis. And in certain ill-informed circles (like the cabal that hands out the fake Nobel prize in economics) it apparently remains so.
In the early days of macrofinance, efforts to isolate a signal for risk appetite focused on the balance sheets of securities broker-dealers. Adrian and Shin argued that dealer leverage, which they had shown was procyclical and the key to the mid-2000s boom-bust cycle in housing finance, was the best proxy of risk appetite. Erkko Etula, who taught me how to be a real quant, developed a measure of the effective risk aversion of the dealers in his doctoral dissertation at Harvard. Zhiguo He and coauthors argued that the capital ratio of the holding company that owns the broker-dealer firm was the best proxy of intermediary risk constraints. I argued that the growth rate of balance sheet capacity itself (whether due to leverage growth or shocks to dealer net worth) contains the strongest signal of risk appetite.
The problem with all these measures is that they’re only available at the quarterly frequency, and that too with an announcement lag. In other words, these measures are practically useless if you want to trade the signal. This has led to me seek other proxies of risk appetite. I have identified two such signals before the one I will introduce today.
First, I found that an inversion of the VIX futures curve (called “backwardation”) predicts risk-offs. The reason is that, because their main risk is shocks to systematic volatility and VIX futures are the price of insurance against shocks to systematic volatility, when dealers’ risk constraints start to bite, they bid up the short end of the curve (up to two weeks out). So, in a true risk-off, the VIX futures curve always inverts. The slope of the VIX futures curve thus contains a strong signal of risk appetite.
Second, following Michael Howell’s work, I found an extremely strong signal in the aggregate equity allocation of US investors. Basically, when risk appetite is high, investors rotate into risk assets; when risk appetite is low, they rotate away from risk assets. Unfortunately, while this signal is fantastic, it is only available at the monthly frequency with a one month announcement lag. From a peak of 71.2% last summer, investors equity allocation has fallen to 61.6% as of October 2022, suggesting a significant if still moderate tightening of financial conditions.
Over the past few years, I have learned a great deal about systematic fluctuations in blue-chip US equities. More recently, I have figured out that one can extract a signal for risk appetite by combining the Asness momentum trick with sector portfolios. In order to do this, we first construct broad sector portfolios of blue-chip US equities. We define momentum as the 230 trading day mean return; lagged by 22 trading days to evade the short-term reversal. Then we compute a sector-blind momentum portfolio as follows. On each monthly rebalance day, we construct an equiweighted portfolio of 200 stocks with the highest momentum. With a Sharpe ratio of 0.64, this is good portfolio; not a stellar one. (Our flagship absolute alpha portfolio has a Sharpe ratio close to 3.) But it can help us isolate a very strong signal of risk appetite.
In order to do that, we look at the weight the momentum portfolio assigns to stocks in the consumer cyclical and consumer defensive sectors. The correlated fluctuations in these weights contains the signal we’re interested in.
Finally, we compute the difference between cyclical and defensive allocations. We call it relative sector momentum, or risk appetite. Simply put, it captures the relative momentum of cyclical vs defensive stocks. We expect it to track risk appetite quite closely. And indeed it does. For instance, in a month where risk appetite is high (in the top third), the expected return on the SPY is 1.3%. In months where risk appetite is low (in the bottom third), the SPY has returned -0.1% on average. In months where risk appetite is moderate (in the middle third), the SPY has returned 0.8%.
The following cross-tab shows the distribution of SPY returns conditional on risk appetite quantile (normalized so that the rows add up to one). Big loss is SPY losing 5% or more in a calendar month; modest loss is a loss of 0-5%. We can see that the probability of a positive return on the SPY rises from 56% when risk appetite is low to 72% when risk appetite is high. Even more strikingly, the probability that the SPY will lose more than 5% in any given month is just 5% if risk appetite is high but 20% if risk appetite is low. So, the probability of market sell-offs rises dramatically when risk appetite is low.
The graph of risk appetite reveals multiple episodes of elevated and subdued risk appetite, with the latter often coinciding with recessions. By the time the NBER declared that the economy was in recession in March 2001, our signal had already turned.
The Great Recession period bears a closer look. According to our signal, risk appetite began collapsing in early 2008 and remained subdued for more than a year. It was not until August 2009 that risk appetite began to rise again.
The episode of Yellen’s failed hiking cycle is also very interesting. Recall our tight money index from the previous dispatch.
Forward guidance on premature tightening from the Yellen Fed led to a collapse of risk appetite in 2016. The market was basically priced for a Fed error-induced recession. Before, of course, ‘the Trump reflation trade’ revived animal spirits. If you recall that November night, the reflation trade began overnight in a very dramatic fashion.
The more recent period is the freshest in our minds. In the quarters prior to the Covid shock, risk appetite was weakening. This is when the yield curve inverted — incredibly (!) predicting the coming pandemic-induced recession. Soon though, the combined fiscal and monetary stimulus led to a roaring revival of risk appetite, which remained elevated until the end of last year. Then it fell through 2022, and is back at ‘recession is imminent’-levels again.
An important question for students of macrofinance is whether there is predictive information about real activity contained in risk appetite. We obtain real GDP from Fred. We resample everything to the quarterly frequency and include a lagged response term to get rid of autocorrelation. We find that risk appetite, as captured by our measure, predicts real activity three quarters ahead. This is consistent with estimates from the yield curve. The reason for this lag is that it takes about that long for persistent shocks to risk appetite to work their way through various credit and risk-taking channels, and reach investment and economic activity.
Our model predicts that real economic growth in 2023 will be no more than half of what it was in 2022. So, we’re definitely looking at recession pricing here.
Note that this measure of risk appetite is extremely persistent. The AR(1) parameter is 0.98 (t=382.0) at the daily frequency, 0.97 (t=133.3) at the weekly frequency, 0.92 (t=35.5) at the monthly frequency, and 0.82 (t=11.9) at the quarterly frequency. Only at the annual frequency does the persistence parameter fall into insignificance. What these means is that our measure of risk appetite, while it is available at arbitrary frequencies, actually captures slow-moving fluctuations in financial conditions that are of relevance to both sophisticated investors and monetary policymakers.
What our measure shows is that risk appetite has collapsed to very low levels. And while there was a glimmer of hope from the inflation print, there is still ways to go before the Fed can let up. Markets are pricing in another 100 basis points or so of further hikes. However, we’re probably getting closer to a pivot, incoming data permitting. Markets will, of course, try to anticipate the Fed’s pivot before it happens. But even so risk appetite may not revive properly for awhile longer if by the time the Fed pivots, the economy is already plunging into a recession. So, it’s not clear how long the present market winter will last. Perhaps it will persist until the exit from the recession is in sight.
You can download our measure of risk appetite from here.
A New Measure of Risk Appetite
Would you be open to sharing the python code for the risk appetite indicator in the dropbox?
Outstanding, as always, Anusar. It feels like we are in a rotational, or rolling recession, where it doesn't hit the economy broadly, but rolls in pockets, similar to a rolling bear market. While services (like travel and leisure) remain strong, the demand for and manufacture of goods has been weakening (evidenced by rising inventories and slowing product sales). This likely recession is rolling through different parts of the global economy at different times, in contrast to the everything-everywhere-all-at-once recessions of 2020 and 2008-2009. You’ve broke down the opportunities here, under these circumstance, elegantly.