Here is the CMI model signal comparison with the US recession forecasts made by IMF, World Bank), and the US Federal Reserve.
Let us consider the evolution of the US recession forecasts made by the International Monetary Fund (IMF), World Bank (WB), and the US Federal Reserve (Fed) from January to October 2008, and the effect of these forecasts on the Fed’s monetary policy. For comparison, the CMI model signal of the US 2007-2009 recession beginning was generated in the summer of 2007 (see “Empirical Verification of the CMI model” section).
January 2008
.According to economic data that was available in January, it was impossible to identify unambiguously that the US economy already was in recession for almost two months (that became officially available almost one year later, i.e., post-factum)[1]. Moreover, available information allowed IMF, WB, and Fed to forecast economic growth both for 2008 and 2009. For example, WB forecasted a +1,9% growth rate for 2008 and +2,3% for 2009[2]. Also, IMF did not expect the recession and forecasted economic growth for the US of +1,5% in 2008[3]. At the same time, Fed forecasted a growth rate of +1,3÷2,0% in 2008 with further acceleration up to +2,1÷2,7% in 2009 and even up to +2,5÷3,0% in 2010[4].
March 2008
The downward trend for financial indexes (that started at the end of 2007) reached its trough in March 2008. It caused some bankrupts for well-known financial banks (Bear Stearns, for example)[5]. The financial shock of March 2008 was accompanied by an economic growth slowdown that caused recession fears, and there was no model to forecast it unambiguously. As result (almost simultaneously) Fed, IMF, and WB downgraded their January forecasts, assuming “mild recession” for the US economy in 2008. Thus, Fed decreased the GDP growth rate forecast for 2008 from 1,3÷2,0% to 0,3÷1,2%, but at the same time increased its growth forecast for 2009 and 2010 from 2,5÷3,0% to 2,6÷3,1% (so, they expected really “mild recession”, probably, as it was in 2001)[6]. And IMF decreased its GDP growth forecast for 2008 from +1,5% to +0,6%[7].
Then Fed and other regulators started acting according to their new forecast: interest rates decreased aggressively from 5% to 2% during 2nd quarter[8] with reduced taxes at the same time[9]. However, regulators had no reliable information about the efficiency of these actions: were the actions enough to avoid a new recession? They were forced to wait for new statistical data to answer this question unambiguously.
June 2008
New statistical data has shown that aggressive regulation policy caused some improvement in the economy. And Fed has stopped the interest rate decreasing fearing inflation accelerating (while crude oil and food prices have been dramatically rising) and declared that inflation (not growth) became the main goal for Fed[10]. After some statistics improved Fed, IMF again increased their GDP forecast for 2008 from 0,3÷1,2% to 1,0÷1,6% and remained unchanged forecast for 2009 (2,6÷3,1%)9,[11]. However, the question — is whether the aggressive policy of regulators was enough to avoid a new recession. — remained without an answer. As became known post-factum only, despite the temporary improvement of the statistics in the 2nd quarter of 2008, new the US recession (later named as Great Recession) was continuing for an 8th straight month already. And 2-3 months remained before the huge financial crash (bankruptcy of Lehman Brothers[12] – one of the biggest the US investment bank).
Thus, only after the Great Recession became evident for everybody did, Fed starts in October 2008 a new highly aggressive monetary policy (named “quantitative easing”) to fight the recession. At least indirectly, it confirms conceptual problems of well-known models to forecast unambiguously the recession starting point. Let us demonstrate these problems by considering some popular practice models as examples.
Composite Leading Indicator Index (CLI-index model)[13]. Figure 1 shows the recession signals of the CLI-index model presented as a probability of the recession. If we assume the threshold value of the recession signal as 0,5, the model will generate two false signals (in 1967, 1996) and missed signal in 2001. If we assume any signal is a recession, about 50% of all signals will be false. Thus, the CLI-index model does not allow us to separate a true signal from a false one for every period.
Fig. 1. Recession probability in the US based on Composite Leading Indicators Index (CLI) (Travis, 2011). Grey columns – official duration of the US recessions (www.nber.org)
Probit-model
Figure 2 presents one of many modifications of Probit-models, i.e. Probit-model with the T-bonds yield spread by Wright (2006)[14]. As Probit-model is some combination of the CLI-index and one financial indicator, it provides better results for recession forecasting than just the CLI-index method.
Fig. 2 Recession signals for the US economy based on the Probit model by Wright, 2006
Grey columns – official duration of the US recessions (www.nber.org)
However, forecasting accuracy of different modifications of Probit-models highly depends on indicators that were chosen to characterize the financial sector of the US economy. Depending on the indicator choice, one version of the model generated the recession signal with a probability of 30% in 2016, but the other one — with a probability of 81% (Ergungor, 2016)[15].
Chicago Fed National Activity Index (CFNAI-MA3)
Figure 3 shows the monthly dynamics of the CFNAI-MA3 index proposed by the Chicago Federal Reserve Bank to forecast the US recessions[16]. However, all the recession signals (within the threshold intervals +0,2 ÷ -0,7) coincide with the official dating of the recessions. It limits the efficiency of regulation and business when this model is used.
Fig.3 Monthly dynamics for CFNAI-MA3 Index for the U.S. economy (Chicago Fed, 2018).
Grey columns – official duration of the US recessions (www.nber.org)
Yield Curve Inversion Method
As it was found empirically, the yield curve spread (or spread between 10- and 1-year of the US T-Bonds or between 10- and 2-year of the US T-Bonds) may serve as a reliable signal for future recessions. Fig.4 shows that the yield spread was negative just before all the US recessions since 1955. However, this model (as any model without theory) has many signals that are difficult to interpret unambiguously.
Fig.4 Monthly dynamics for yield spread: 1) between 10- and 1-year of the US T-Bonds (solid line);
2) between 10- and 2-year of the US T-Bonds (dotted line) Grey columns – official duration of the US recessions (www.nber.org)
Source: (Federal Reserve Bank of St Louis, https://fred.stlouisfed.org).
First, we can see the false signal of 1966 that is difficult to explain. Second, before 1980-th the recession deepening was coinciding with the yield curve inversion deepening (its bottom was coinciding with the trough of the recession). But the yield curve inversion was appearing and disappeared before starting points for the last three recessions (after 1980-th). We can explain it by changing the main principles of the Fed’s monetary policy (from Keynesian principles before 1980-th to monetarist ones after them), which affect the interest rates and bond yield directly. However, it remains unclear: how does future monetary policy can affect the yield curve inversion?[17] Will the yield curve inversion appear after “quantitative easing” monetary policy (QE)?
Some researchers suppose that the yield curve method will not lose the ability to forecast a recession even after an unusual monetary policy (QE)[18]. Others believe that this inversion would not appear at all, and, if it even appears, the new recession will not happen[19]. H. Yellen supported this thought in her last speech as Fed Chairman. She supposes that the yield curve inversion may appear after tightening monetary policy, but it may not be a signal for a new recession[20].
Any generalization with the yield curve inversion method becomes complicated by the fact that this method may be more or less accurately used to forecast recessions for the US economy. However, it loses its accuracy significantly when the recession forecasting for the German or British economy[21].
Thus, we can summarize the common drawbacks of well-known models, as noted above:
1) No one model is general. Every model is valid for separate (specific) economies and
corresponding market conditions.
2) Every model generates false or/and missed signals.
3) If the signal is unambiguous, it will be generated too late to make regulation or business effective.
All the drawbacks provide permanent ambiguity and incompleteness of information used for decision-making.
References
[1] US National Bureau of Economic Research: US Business Cycle Expansions and Contractions (Recessions) —
Retried from: www.nber.org
[2] Global Economic Prospects. (2008) Technology Diffusion in the Developing World. The International Bank for
Reconstruction and Development / The World Bank. — Retried from:
http://siteresources.worldbank.org/INTGEP2008/Resources/complete-report.pdf
[3] Davis B. (2008) IMF Predicts Slower U.S. Growth With Global Drag // The Wall Street Journal, January 30, 2008
[4] Board of Governors of the Federal Reserve System (2008). Minutes of the Federal Open Market Committee,
January 29-30, 2008.— Retried from: https://www.federalreserve.gov/monetarypolicy/fomcminutes20080130ep.htm
[5] Bear Stearns — New-York investment bank (founded May1, 1923).
[6] Reddy S. (2008) Fed Holds Rate Steady as Inflation Worries Rise // The Wall Street Journal, June 26, 2008.
[7] Barkley T, Hannon P., Chalton E. (2008) IMF Sees U.S. Recession, Slowing Global Growth // The Wall Street
Journal, April, 2008
[8] Board of Governors of the Federal Reserve System (2008). Minutes of the Federal Open Market Committee, June 24-25, 2008. — (https://www.federalreserve.gov/monetarypolicy/fomcminutes20080625ep.htm)
[9] Bater J. (2008) Tax Rebates Widen U.S. Deficit // The Wall Street Journal, June 11.
[10] Inflation Now Enemy No. 1 for Fed // The Wall Street Journal, June 19, 2008
[11] Barkley T. (2008) IMF Raises Global Growth Forecast // The Wall Street Journal, July 18, 2008
[12] Lehman Brothers — 4-th in size the US investment bank founded in 1850.
[13] Travis J, Berge T., Elias E., Jorda O. (2011) Future Recession Risks: An Update // Federal Reserve Bank of San Francisco Economic Letter. – No35. – Pp.1-15.
[14] Wright, J.H., (2006). The Yield Curve and Predicting Recessions. Finance and Economics Discussion Series, Federal Reserve Board, February.
[15] Ergungor E. (2016) Recession probabilities // Economic Commentary No 2016-09. Federal Reserve Bank of Cleveland. — p.6 — Retried from: (www.clevelandfed.org/research)
[16] Chicago Fed National Activity Index (CFNAI-MA3) Monthly Release (2018) // Federal Reserve Bank of Chicago. — Retried from: (https://www.chicagofed.org/research/data/cfnai/current-data)
[17] Bauer M, Martens T. (2018) Information in the Yield Curve about future recessions // FRBSF Economic Letter 2018-20 (August 27). Federal Reserve Bank of San Francisco — Retried from (https://www.frbsf.org/economic-research/publications/economic-letter/2018/august/information-in-yield-curve-about-future-recessions/ )
[18] Bauer M, Martens T.(2018) Economic forecasts with the Yield Curve // FRBSF Economic Letter 2018-07 (March 5). Federal Reserve Bank of San Francisco — Retried from:
[19] Christensen J. (2018) The slope of the Yield Curve and near-term outlook // FRBSF Economic Letter 2018-23 (October 15). Federal Reserve Bank of San Francisco — Retried from: (https://www.frbsf.org/economic-research/publications/economic-letter/2018/october/slope-of-yield-curve-and-near-term-outlook/)
[20] Chappatta B. (2017) Yellen tells investors not to fear the flattering yield curve // Bloomberg. Market News.
December 14, 2017 — Retried from:
[21] Wheelock D., Wohar M. (2009) Can the term spread predict output growth and recessions? A survey of the literature // Federal Reserve Bank of St. Lois Review. September/October 2009, 91(5, Part 1). —pp.419-440.