Monday, 30 June 2008

The Greenspan Put

As an alternative to the savings glut explanation for the expansion of spread product in the US that ended in the sub-prime bust, in his previous post RebelEconomist suggested that investors’ belief in the existence of the so-called Greenspan Put may have played a role in encouraging them to buy more risky assets. This post presents some evidence that investors do seem to have come to expect the Fed to at least behave as if there is a Greenspan Put.

The idea of the Greenspan Put is that the Federal Reserve provides protection for investors against sharp falls in the value of US risky assets (strictly, a put option hedge would be paid for and exercised by the investor, but it is the protection principle that matters here). By cutting interest rates in response to such developments, the Fed can support asset prices, both directly by lowering the discount rate applied to anticipated future cashflows, and indirectly by boosting economic activity which decreases the likelihood of debt defaults and, in the case of stocks, enhances company earnings. While it can be argued that easing monetary policy is an appropriate reaction to a large fall in asset prices anyway, because such events inhibit economic activity and hence restrain inflation, under Alan Greenspan's leadership the Fed made a series of eases in response to financial crises which appeared to be disproportionately aggressive considering either the severity of the shock or the amount of easing in each case.

Within a few weeks of Greenspan taking over from Paul Volcker as Fed chairman, the Fed cut the Federal funds target by half a percent to bolster damaged confidence in the wake of the “Black Monday” stock market plunge of October 19th 1987, even though this roughly 25% correction merely cancelled out the increase of the preceding year of boom. Fed funds were held at 3% from September 1992 to February 1994 to assist the resolution of the “thrifts crisis”. Interest rates were cut when Long Term Capital Management’s disastrous bets on continued convergence of yield spreads caused trouble in Autumn 1998, despite the fact that Greenspan had specifically warned that “caution also seems warranted by the narrow yield spreads” in his Humphrey Hawkins testimony of February 1997. Greenspan subsequently explained that the easing was intended “to address the seizing-up of financial markets”. This cut in interest rates, and the various liquidity-ensuring measures taken in case of disruption associated with the turn of the millennium, seemed to accelerate the persistent rise of the stock market from the mid-1990s that had prompted Greenspan’s famous “irrational exuberance” remark. Nevertheless, when the dotcom boom turned to bust, the Fed reduced the Fed funds rate to a historic low, though the subsequent recession turned out to be mild and inflation remained comfortably positive. Although Greenspan retired in January 2006, the rapid lowering of interest rates by the Fed since the current financial crisis began with losses on US sub-prime mortgage debt, especially the apparently panicky 75bp emergency cut on January 22nd this year, suggests that the pattern continues with new chairman Ben Bernanke, and there is now discussion of a Greenspan-Bernanke put.

As Figure 1 shows, these episodes seemed to change the link between interest rate expectations and risky asset prices. If the central bank is mainly concerned with inflation, then interest rate expectations will tend to drive risky asset prices, and bond yields which discount short-term interest rates will tend to be negatively correlated with stock prices, because, other things equal, higher interest rates are bad for risky assets and low interest rates good. Alternatively, the more the central bank is inclined to ease to mitigate falling asset prices, the more frequently asset prices drive interest rate expectations, in which case the correlation between short-term interest rates and asset prices will be positive. Figure 1 shows the correlation between daily changes in yield on the two year treasury note and in the S&P500 stock index from August 6th 1979, the day that Volcker became Fed chairman, until June 26th 2008. The two year note yield tends to be highly sensitive to near-term policy interest rate expectations without, as a “risk-free” rate, being disturbed by changes in creditworthiness. The S&P500 index represents the main risky asset class.

In the early years of the period covered by Figure 1, the blue line, which shows the rolling one hundred (working) day correlation between interest rate and stock price changes, indicates generally negative correlation. Its first sustained move into positive territory is linked with the 1987 stock market crash - note that, as each point in the hundred day sample contributes equally to the correlation and the correlation value is plotted at the point corresponding to the centre of the sample, sharp changes in market behaviour will appear fifty days before the event that drives them, and last for fifty days afterwards. Stock market sentiment was fragile for several years after the 1987 crash, and there was a brief scare when stocks fell sharply again on Friday 13th October 1989, which again drove correlation positive, although the Fed did not actually ease on that occasion. By going ahead with tightening in February 1994 despite being doubted by the market, causing considerable stress in fixed income markets, Greenspan acquired a more hawkish reputation, and the correlation reached its most negative in the summer of 1994. The Fed’s response to the 1998 LTCM crisis, however, generated a break in market expectations, since when the correlation has been mainly positive, and presently stands only just below the high reached in late 2002 in the stressful period after the 9/11 attacks and leading up to the Iraq War.

The red and green lines show the correlation for the Volcker and Greenspan-Bernanke eras respectively (practically the average of the blue line within each sub-period). Under Volcker, a celebrated inflation-fighter, the correlation was clearly negative (-0.283) and much lower than during the tenures of Greenspan and Bernanke so far (+0.065). Shifting the interest rate and stock price changes by one day sheds additional light on market behaviour during the two regimes. In all cases, the correlation between the shifted series is nearer zero, but under Volcker the negative correlation shrinks by far the most when stock price changes are shifted one day forward, whereas under Greenspan the positive correlation is only slightly lower when stock prices are shifted forward one day. This is consistent with the theory that monetary policy tended to drive stock prices in the Volcker era, whereas stock prices have more often driven monetary policy expectations since then.

In conclusion, while not necessarily responding to risky asset price changes per se (because, say, the Fed governors wish to be popular), the fact that the Fed has eased during times of market turmoil against a background of generally stable economic activity and comfortably positive inflation has conditioned the market to expect the Fed to any mitigate sharp fall in risky asset prices. The danger is that investors have come to rely on such a response and therefore allocate more of their wealth to risky assets. Then, if the Fed fails to live up to investors’ expectations following some sharp fall in risky asset prices, that fall will be exacerbated as expectations are revised. In short, the Fed seems to have created a moral hazard trap. The present conjuncture of financial market turmoil, weakening US economic activity and rising inflation may well bring matters to a head. So far, while investors have begun to recoil from particular types of spread product, such as asset-backed commercial paper (ABCP), their continued inclination to respond to savings shortfalls by increasing exposure to risky assets rather than saving more suggests that investors continue to rely on the Greenspan Put. Either the Fed will allow higher inflation or investors in risky assets are going to be disappointed.


In response to a good question from an anonymous commenter, I checked that the two era correlations are statistically significant from zero. Despite the fact that the correlations are not large, they are highly statistically significant, because of the huge number of daily observations in each era. There are 2026 points in my Volker era and 5266 in my Greenspan/Bernanke era, so the correlations of -0.283 and 0.065 have z-scores of -12.7 and 4.7 respectively. These are both statistically significant from zero even at even the most stringent level like 0.1% (two-tailed). Actually, it is arguably the difference between them that matters, which is of course even more significant since the two correlations have opposite sign.

For anyone who is interested in the technical details of this analysis, I transformed the correlations using Fisher's z-transformation. This yields a variable that is approximately normally distributed with known mean and standard deviation, which can then be checked for significance using tables for the standard normal distribution.


Anonymous said...

Not clear that a correlation of 0.065 is statistically different from 0.

RebelEconomist said...

Anonymous, you raise a good question. Even this level of correlation is highly statistically significantly different from zero because of the huge number of daily observations. There are 2026 points in my Volker era and 5266 in my Greenspan/Bernanke era, so the correlations of -0.283 and 0.065 have z-scores of -12.7 and 4.7 respectively. These are both statistically significant from zero even at even the most stringent level like 0.1% (two-tailed), although it is arguably the difference between then that matters, which is of course even more significant since the two correlations have opposite sign.

Anonymous said...

Many thanks for responding.

Anonymous said...