Different forward looking models show increasing likelihood of a recession. Most recent readings of key series highlighted by the NBER’s Business Cycle Dating Committee (BCDC) suggest a peak, although the critical indicator — nonfarm payroll employment — continues to rise, albeit slowly.
Predictions from Financial Markets
Figure 1: Probability of recession in indicated month, based on spreads 12 months prior, from 10yr-3mo term spread (blue), 10yr-3mo term spread and Moody’s Baa-10yr credit spread (red), 10yr-3mo term spread adjusted by ACM 10yr term premium (teal), and 10yr-3mo term spread and ACM 10yr term premium (black). Probit regressions estimated on monthly data over 1986-2019M05 data. NBER defined recession dates shaded gray. Source: FRED data (including NY Fed via FRED), NBER, and author’s calculations.
It’s definitely true that the premium adjusted term spread predicts a very low probability (teal line); however using a 33% threshold, that model would have missed completely the 2007-09 recession (i.e., false negative). Of course, using the 50% threshold no recession is imminent.
Note that using a 33% threshold and the simple term spread model (blue line) would suggest a recession in 2020, while not sending a false positive for late 1999. The term and credit spread model (red line), while fitting well, would’ve provided a false positive for late 1999 using that same 33% threshold. (A good recent survey and evaluation of competing spreada models is provided by David Miller in FEDS Notes.)
Figure 2: Nonfarm payroll employment (blue), industrial production (red), personal income excluding transfers in Ch.2012$ (green), manufacturing and trade sales in Ch.2012$ (black), and monthly GDP in Ch.2012$ (pink bold), all log normalized to 2019M01=0. Source: BLS, Federal Reserve, BEA, via FRED, Macroeconomic Advisers (5/30 release), and author’s calculations.
Nowcasts extending out to 2019Q3 all suggest a slowdown (as indicated in flattening trajectories in GDP in Figure 3), but not necessarily a recession (no negative entries in Figure 4).
Figure 3: GDP in bn.Ch.2012$ SAAR as reported in second release (black), Macroeconomic Advisers of 6/6/2019 (blue), NY Fed Nowcast of 6/7/2019 (red), and Atlanta Fed of 6/6/2019 (teal). Source: BEA, NY Fed, Atlanta Fed, Macroeconomic Advisers, and author’s calculations.
Figure 4: Annualized quarter on quarter growth rate of real GDP as reported in second release (black), Macroeconomic Advisers of 6/6/2019 (blue), NY Fed Nowcast of 6/7/2019 (red), and Atlanta Fed of 6/6/2019 (teal). All calculated using log differences. Source: BEA, NY Fed, Atlanta Fed, Macroeconomic Advisers, and author’s calculations.
Expectations/Uncertainty Sensitive Components of Aggregate Demand
All economic activity depends on expectations of future variables — including asset prices, which is why the expectations hypothesis of the term structure is useful. However, some components of aggregate demand are more sensitive to expectations — both mean, variance, and uncertainty — than others. Investment — particularly equipment — is an obvious example; so too is housing investment.
Figure 5: Equipment investment (dark blue), and capital goods imports excluding aircraft and computers (red), four quarter change as log ratio to GDP, all in 2012Ch.$. NBER defined recession dates shaded gray. Light orange denotes Trump administration. Orange denotes TCJA in effect for equipment. Source: BEA 2019Q1 2nd release, NBER, and author’s calculations.
Equipment investment is now flat y/y. Imports of capital goods aside from computers and aircraft and aircraft parts are now decreasing, although it’s hard to say how much of this is due to tariffs. While both series predict recessions (McFadden Rsquared equal to 0.36), the implied recession probability is only about 5% in one year (see this post).
Figure 6: Investment in single family residential housing, in bn.Ch.2012$ SAAR (blue, left log scale), and Case-Shiller 20 city house price index, deflated by CPI-All (red, right log scale). NBER defined recession dates shaded gray. Light orange denotes Trump administration. Orange denotes TCJA in effect for investment. Source: BEA 2019Q1 2nd release, S&P and BLS via FRED, NBER, and author’s calculations.
What about Leamer’s thesis of the business cycle as housing cycle? In the last recession, investment and real house prices peaked about two years before the onset of the recession. Real investment peaked in 2018Q1. Real housing prices, on the other hand, have not yet peaked.
The Case-Shiller indices only go up through March. Looking at the monthly Zillow price indices, which extend through April, one can see an apparent peak: March for nominal, February for real.
Figure 7: Zillow median single family house price, in $ (blue), and in 2018$ (red). Real calculated using CPI-All. Source: Zillow, BLS via FRED, NBER, and author’s calculations.
Moreover, in the top 12 home markets, 11 are experiencing decreasing prices from March to April (the largest — New York city — is flat).
Interestingly, the Zillow index (which differs from the Case-Shiller, see discussion here) peaks one year before the recession. If we get a repeat performance, the recession begins in February 2020.