Global Economics Analyst
Recession Risk is Low…For Now
Table of Contents
We expect strong global growth this year, given firm current momentum, easing financial conditions, and supportive fiscal policy. But high asset valuations and the prospect of labor market overheating suggest that the recent strength might be “too much of a good thing” further down the road.
We revamp our cross-country recession model to gauge the risk of a downturn across major advanced economies with our proprietary indicators. At short horizons of less than a year, weak growth momentum (as measured with our CAIs) and tightening financial conditions are the best recession predictors. At longer horizons, output above potential, and unusually accommodative FCI levels signal recession risk.
Our model suggests that near-term recession risk is low. The probability of a downturn is also below normal over the next 2-3 years, but has been rising steadily in economies that are seeing unusually easy financial conditions and tightening labor markets. These include the US, Germany, the UK and a number of smaller G10 economies (such as Canada and Sweden). Medium-term recession risk, in contrast, remains subdued in countries with remaining slack, including Spain, Italy and France.
Although our model is subject to a number of caveats, it confirms that we need to worry little about recession risk this year. But our analysis suggests that we should pay attention to measures of imbalances that signal rising recession risk further down the road. Our new recession model provides a tool to track these risks in real time.
The world economy has continued to grow very strongly. Our global current activity indicator (CAI) stands at 5.1% in December, after 5.6% in November. Financial conditions have continued to ease, providing a continued boost to growth across the globe (Exhibit 1). Fiscal policy has also turned expansionary with the US tax cuts, boosting growth by an estimated 0.4 percentage points (pp) across the large advanced economies this year. Given strong momentum and positive growth impulses from financial conditions and fiscal policy in advanced economies, we are optimistic about the world economy’s outlook and expect global growth of 4.1% this year, with risks that still seem skewed to the high side.
Exhibit 1: Financial Conditions Have Continued to Ease
Source: Goldman Sachs Global Investment Research
But early signs of overheating are beginning to emerge in a number of advanced economies. Although inflation has generally remained low, our analysis suggests that slack is diminishing rapidly across DM economies. We estimate that the US, Germany, the UK and a number of smaller G10s (including Canada, Sweden and New Zealand) have already moved past full employment (Exhibit 2). The global risk rally—with the S&P hitting new all-time highs day after day—has pushed our FCIs to unusually accommodative levels. Our global FCI, for example, is now at the most accommodative level since mid-2013. Overheating labor markets and stretching valuations could soon be “too much of a good thing,” and might raise the probability of a risk further down the road.
Exhibit 2: Slack is Diminishing
Source: Goldman Sachs Global Investment Research
A New Cross-Country Recession Model
We revisit our global recession model to gauge the risk of recession across the advanced economies.[1] Our revamped model is similar in spirit to our previous approach in using economic and financial variables to predict recessions across major advanced economies. Using cross-country information adds statistical power (as recessions are rare events) and provides us with estimated recession probabilities for multiple economies. We again exclude emerging markets, which typically feature very different business cycles from the advanced economies.
But our new model differs from our previous work in a number of dimensions.
First, we put greater emphasis on our proprietary indicators to gauge recession risk. We include four domestic explanatory variables : (1) growth momentum, measured with our CAIs relative to our estimate of potential growth; (2) slack, measured with our output gap estimates;[2] (3) changes in financial conditions, measured with the year-over-year change in our FCIs; and (4) FCI levels, expressed in deviations from a five-year centered moving average. Using our proprietary indices has the advantage that they are much more timely than national-source data such as GDP, and this helps us monitor recession risk at a higher frequency.
Second, we now use the recession definitions published by the Economic Cycle Research Institute (ECRI). The ECRI follows the methodology of the US National Bureau of Economic Research (NBER) in timing troughs and peaks of the economic cycle. Exhibit 3 shows the implied number of recessions in our sample.
Exhibit 3: Counting Recessions
Source: Haver Analytics, Goldman Sachs Global Investment Research
Using these recession quarters, we construct recession “alarm” indicators for different horizons, including the current quarter, the next year, the next two years and the next three years (all of which include the current quarter). For example, the “next year” indicator would cover the incidence of a recession in 2018.[3]
Exhibit 4 shows the implied unconditional probabilities of being in recession in any given quarter since 1980. The main take-away is that unconditional recession probabilities are sizable. For the US economy, for example, the unconditional probability of being in recession is 14% in any given quarter, and as high as 31% over any given two years (not shown). Moreover, unconditional recession probabilities vary significantly across countries. For example, Italy and Japan have had high average recession risk—partly because of low potential growth—whereas the US and Australia have had relatively low recession risk.
Exhibit 4: Unconditional Recession Risk Across Countries
Source: Goldman Sachs Global Investment Research
Third, we explicitly take into account the cross-country correlation of recession risk. Exhibit 5 shows that recessions are generally correlated positively across advanced economies. The correlation pattern, however, differs across countries: recession risk in Sweden, for example, appears highly correlated with foreign downturns, while the correlation is much lower in Japan. We try to capture this pattern by including the export-weighted foreign output gap in each country’s recession model. Moreover, we interact the foreign output gap with the trade openness of the home country in an attempt to capture the cross-country pattern of correlations in Exhibit 5.
Exhibit 5: Recessions are Correlated across Countries
Source: Haver Analytics, Goldman Sachs Global Investment Research
Fourth, we now use simpler statistical techniques to estimate our recession model (replacing the “probit” specification with a linear probability model). The advantage of our new approach is that the estimated coefficients are much easier to interpret.[4] We include country-specific “dummy” variables to capture local average recession risk. We estimate our model using quarterly data for 13 advanced countries since the mid-1970s.
Predictors of Recession
Exhibit 6 shows our results. The first row for each variable in the table shows its economic importance and the second row shows the statistical significance (summarized with the t-statistic and asterisks that denote statistical significance).
Exhibit 6: Our Estimated Model
Source: Goldman Sachs Global Investment Research
Our main results are as follows.[5] First, the state of the domestic cycle matters. Above-trend growth lowers recession risk while overheating raises the chance of a downturn. The effect of current growth momentum is stronger for the risk of near-term recession; but the output gap matters more for future recessions. For example, one-year recession risk is just over 3½pp lower for every one percentage point that the CAI exceeds trend. But two-year recession risk is 2pp higher for every one percentage point of overheating today.
Second, the change in financial conditions is a powerful recession predictor. Specifically, we find that FCI easing lowers recession risk. Thus, while the markets frequently sound false alarms, ignoring them entirely would be a grave mistake. For example, a 100bp easing in financial conditions over the last year lowers the chance of one-year-ahead recession by just over 2pp. The year-over-year change in the FCI closely resembles our FCI impulse.
Third, the level of financial conditions holds additional information. Accommodative financial conditions (relative to a moving average) suggest that recession risk is typically elevated. This effect operates through recession risk further out: if financial conditions are 100bp easier, the three-year recession probability rises by more than 8pp. In effect, one might say that our FCI is a financial analogue to the output gap: near-term strength reduces the chance of a downturn but a very strong level (relative to the long-term average) indicates a potential overheating that must be reversed, with adverse effects on financial markets and the real economy.
Fourth, foreign overheating matters for domestic recession risk. We find that the trade-weighted foreign output gap has a small, but statistically significant effect on domestic recession risk. Moreover, the strength of this relationship varies with the openness to trade; for example, German recession risk depends much more on the foreign cycle than US recession risk does.
We note that neither the age of the recovery nor the depth of the previous recession matter once the other variables in the model—especially the output gap—are included.
Recession Risk is Low…For Now
Exhibit 7 shows the estimated recession risks for the current quarter, and the next 4, 8, and 12 quarters for all the individual economies in our model. We also include the unconditional risks of recession at each horizon, which differ substantially across countries.[6] We truncate the recession probability estimates outside the 95% range as our confidence in their accuracy is low (showing, for example, <2.5% for very low probabilities).[7] We also provide joint probabilities for a Euro area recession (proxied with the 'big four' member states) and for recession risk across the majority of advanced economies.[8] Exhibit 8 decomposes the current recession probability into contributions from the explanatory variables included in the model.
Exhibit 7: Recession Risk is Low…
Source: Goldman Sachs Global Investment Research
Our model confirms that the risk of near-term recession is low across all advanced economies. For example, the one-year probability of recession is 5% in the US (compared with the unconditional risk of 22%), < 2.5% in the Euro area (versus 33%) and 21% in Japan (versus 44%). The decompositions for the US and Italy in Exhibit 8 show that—in general terms—recession risk is low primarily because growth momentum is strong and FCI impulses are positive. But while slight overheating in the US pushes up the risk in two to three years' time, the large negative output gap in Italy weighs down on risks over the same horizon.
Exhibit 8: …Owing to Strong Growth and Easing FCIs
Source: Goldman Sachs Global Investment Research
Recession risk over the next 2-3 years is also below normal across the advanced economies. But Exhibit 9 shows that the medium-term risk has been rising steadily in the US given easy financial conditions and a tightening labor market. This is similar in the UK, Germany and a number of smaller G10 economies (including Canada and Sweden). Exhibit 10 shows that medium-term recession risk, in contrast, remains subdued in the Euro area, because southern Europe is still operating below capacity.
Exhibit 9: But 2-3 Year Risk is on the Rise in the US…
Source: Haver Analytics, Goldman Sachs Global Investment Research
Exhibit 10: …Although Still Low in the Euro Area
Source: Haver Analytics, Goldman Sachs Global Investment Research
Exhibits 9 and 10 suggest that the fit of the model is good. In general, recession risk was high before and during past recessions but not otherwise.[9] It is, however, important to stress a few caveats. First, the out-of-sample performance of the model may not match the in-sample fit. This is partly because the model is estimated using revised rather than first-release data and, more broadly, because no two recessions are the same. Second, our model does not capture a number of other risks, such as the effects of uncertainty or political instability. As a result, our recession probabilities might be understated for Italy (given the Italian elections) and the UK (in light of Brexit uncertainties).
With these caveats in mind, we conclude that we need to worry little about recession risk this year. But our results suggest that we should pay attention to measures of imbalances that signal rising recession risk further down the road. Our new recession model provides a tool to track these risks in real time.
Nicholas Fawcett
Sven Jari Stehn
Manav Chaudhary
Global Economic Forecasts
Source: Goldman Sachs Global Investment Research
- 1 ^ For details of our earlier model, see Jan Hatzius, Sven Jari Stehn and David Mericle, “Doing the Numbers on DM Recession Risk,” Global Economics Analyst, February 5, 2016.
- 2 ^ For details of our potential growth estimates in DM economies, see Sven Jari Stehn, “From Demand to Supply: Our New G10 Output Gap Estimates,” Global Economics Analyst, November 5, 2017.
- 3 ^ The one-year alarm, for example, is equal to 0 in general, but switches to 1 four quarters before the start of a recession; it remains at 1 until the end of the recession episode. The two-year alarm switches to 1 eight quarters before the start of the recession, and so on.
- 4 ^ The new approach also eliminates the bias in panel probit models with fixed effects.
- 5 ^ If the country was already in recession in the previous quarter, then this raises substantially the probability of being in recession in the current quarter. We do not report this effect in the Exhibit, and focus instead on the most important drivers, assuming that the economy is not already in recession.
- 6 ^ These differences are explained by factors such as the trend rate of growth, and volatility of growth.
- 7 ^ Since we use a linear model, rather than one tailored to probabilities that lie between 0 and 1, it is possible that extreme readings of the explanatory variables lead to implied recession probabilities of less than 0 or greater than 1. In these circumstances the model may not capture the true effect of the drivers on the recession risk. Reassuringly, this happens relatively infrequently, so this problem is not severe in practice.
- 8 ^ If enough countries in an area enter into recession at once, then the area as a whole may be tipped into recession. We consider all the possible combinations of countries which would satisfy this requirement, and use our assessment of risk for each country, to evaluate the risk for the area as a whole.
- 9 ^ This is echoed by more formal comparisons of actual recessions to the predicted risk.
Investors should consider this report as only a single factor in making their investment decision. For Reg AC certification and other important disclosures, see the Disclosure Appendix, or go to www.gs.com/research/hedge.html.