Our investment decision-making process incorporates three key components: business cycle, valuation, and technical analysis. The first in a series on our investment process introduces our Recession Risk Indicator (RRI). We outline how we use the RRI to evaluate the likelihood of economic recessions and examine how the indicator can help us navigate the later stages of the business cycle.
Business Cycle Analysis: Recession Risk Indicator
Given the varying lengths and magnitude of economic cycles, it’s natural to ask, how much time is left in the current expansion? Historically, equity bear markets have coincided with recessions; therefore, we believe it is important to carefully analyze the business cycle for economic risks that may lead to a more protracted market downturn. While we do not believe an economic recession is imminent, several indicators suggest to us that we are now in the later stages of the business cycle.
To better forecast the most challenging business cycle phase, we have developed the Recession Risk Indicator (RRI). The RRI comprises several leading and coincident economic/market indicators that offer various perspectives on the current state of the economy. The RRI uses probit regression to determine the predictive power of each variable, resulting in an indicator output with a defined range between 0 and 100. Periods that generate higher RRI readings are indicative of an environment more prone to recession, while lower readings suggest a lower probability of such an occurrence.
Here we explore the RRI, focusing on the following key insights:
- What is it? Defining the RRI components and methodology.
- How to use it? Interpreting RRI levels.
- How to navigate late-cycle markets? Looking at RRI and bear markets.
- How does the RRI fit within our overall business cycle framework? Using the RRI in combination with our other business cycle analysis.
RRI Components and Methodology
The RRI is derived from 13 individual components categorized as either early or affirming indicators. Early indicators are metrics that have historically led recessions by greater than 12 months while affirming indicators help confirm the trend in the leading metrics, either coinciding with recessions or leading to a lesser extent.
Each individual metric was selected to capture different portions of the economy. Our analysis leads us to believe a diverse combination of variables best captures the unique nature of each economic cycle, helping us to better identify environments most susceptible to a contraction in economic growth. Additionally, we analyzed each indicator based on its variability in the 24 months prior to each historical recession, allowing us to confirm more consistent signals from each RRI component. Economic indicators with less consistent levels leading into a recession are far less useful — in the case of wide variability, averages can be deceiving.
The calibration of recession risk is accomplished through probit regression. Probit regression is a type of regression with a binary dependent variable, which in our case is a recession time series that can take the value of 1 (in a recession) or 0 (not in a recession). The regression is run on each indicator’s historical monthly values to quantify levels most often associated with a recession. The final output for each indicator is a value between 0 and 100.
To help reduce volatility and prevent outliers from skewing results, all indicator values are winsorized to three standard deviations.
A binary dependent variable is a dummy variable that takes on a value of either 0 or 1 depending on an observed result that can have only two possible outcomes. In our regression, the variable is either 0, meaning the economy is not in a recession, or 1, indicating it is in a recession.
Interpreting the Recession Risk Indicator
As we know, timing markets and predicting recessions is difficult. However, we believe our disciplined approach to analyzing the business cycle helps us identify certain conditions that make a recession more probable and therefore could put the equity market at risk. Recessions and bear markets are usually closely linked. Therefore, understanding inflection points within an economic cycle, particularly when the U.S. economy is likely to enter a recession, could be a significant benefit to any asset allocation process. Our RRI was designed with this very purpose in mind.
Understanding the RRI and Subindicators
When interpreting indicator results, it is useful to understand both the overall indicator level and each subcategory level. The overall RRI measures the simple average of the early and affirming subindicators, with a level of 50 as the base value. An indicator reading above 50 and rising signifies a higher likelihood of recession, while levels below 50 indicate a lower likelihood of recession (the same is true for our early and affirming subindicators). Once the RRI level rises above 50, recessions have resulted on average 23 months later.
|Current RRI Level
||Above Avg. Risk (>50)
|Investment as a % of GDP
|Average Hourly Earnings||45.9
|Output Gap % Potential GDP
|Domestic Corporate Profits % GDP
|10-Year Treasury and 3-Month T-Bill Spread
|Consumer Interest Expense
|Nonfinancial Corporate Interest Expense
|Single-family Housing Starts||65.2||Yes|
|Global 10-Year Yield
From February 1980 through November 1982, the United States suffered a double-dip recession, according to the NBER. In our analysis, we treat this period as one long recession. Using the NBER definitions, there were approximately 12 months between the two technical recessions; however, the most watched economic indicators never truly recovered, and neither did financial markets. Overall GDP was negative for 6 of the 12 quarters, and unemployment was above 10% for a significant portion of the time. Additionally, our RRI level never moved back below 50 in the time between the two recessions, hovering around 70 or higher for the entire period.
|Recession Start Date||Early RRI||RRI||Total RRI
|Average Lead Time
*This period included both the 1980 and 1981-82 recessionary periods.
At the category level, the early subindicator, as its name would suggest, has offered additional lead time, with recessions occurring 25 months on average after crossing above 50. Comparatively, the affirming subindicator has crossed over the 50 mark 17 months prior to recession, on average.
As the economy moves closer to recession, the level that would indicate a need for increased caution becomes more subjective. In the month prior to the beginning of a recession, the average reading of the RRI has been 77; the highest, 82; and the lowest, 68.
Early and Affirming Subindicators
The interaction between our early and affirming subindicators is an important, albeit more nuanced, element when interpreting overall RRI levels. Specifically, the two subindicators serve as a checks-and- balance system, helping to uncover possible false-positives. Defensively adjusting asset allocations too early can be just as painful as the ensuing market downturn. We attempt to mitigate the risk of reacting to falsepositives by reviewing how the early and affirming subindicators behave relative to each other moving through the business cycle. For example, the longest false-positive period when looking at the overall RRI occurred in the mid-1990s, when our affirming subindicator pulled the overall RRI above 50 for 12 months before ultimately falling back below 50. Given the expected relationship between early and affirming subindicators (that is, early leads only to be followed by affirming when the economy is approaching recession), we would have been correctly skeptical of the overall RRI signal at that time.
Each Recession Is Different
The NBER defines a recession as “a significant decline in economic activity spread across the economy, lasting more than a few months…” But what causes a recession? Many economists have spent much of their careers attempting to answer this question. The simple answer is, it’s always a different economic or financial component of our evolving economy that ultimately pushes us into an economic contraction.
Our RRI metrics can be plotted by their individual recession indicator levels prior to entering each of the last six recessions — and the primary culprit is always a different variable.
For example, leading into the financial crisis of 2008, the metrics we included to capture changes in global interest rates and leverage levels were both at their highest RRI readings within the context of our analysis. These excesses, in terms of borrowing and ultimately the securitization of that debt, resulted in the collapse of the subprime mortgage market and caused stress across the U.S. banking system, which subsequently pulled the U.S. economy into contraction. Fast forward to today: Lending standards have improved, subprime mortgages are a smaller portion of the overall mortgage market, and banks are well-capitalized.
As another example, the recession of 1982 was primarily driven by a rapid increase in energy prices as a result of the 1979–80 oil output cuts in Iran and Iraq. This led to a rapid rise in inflation, prompting then Federal Reserve Chairman Paul Volcker to rapidly increase short-term interest rates. Today, the economy is still highly sensitive to energy costs, but the United States is much less dependent on foreign oil; in fact, the United States is now a significant oil exporter.
For this reason, we include 13 carefully selected indicators in our recession analysis, equally weighting them given the knowledge that each historical recession looks somewhat unique.
|Year of Recession
||RRI at S&P 500 Peak
||Months from Peak to Recession
||Decline from Peak to Recession
||S&P 500 Peak to Trough Drawdown
|Average Excluding 1973||76.2||3.8||-10%||-40%|
RRI and Bear Markets
History has shown that prolonged equity market downturns are far more likely during recessions. Market declines during expansionary periods have been less severe in both magnitude and duration relative to market drawdowns coinciding with recessions. The RRI was designed to provide insight into recession-driven bear markets where exercising precaution offers the greatest benefit.
Excluding the 1973 period, the S&P 500 has produced an average annualized return of 10% or greater between the RRI reaching 50 and the beginning of the next bear market. Thus, investors risk missing fairly robust return periods if portfolios are adjusted defensively too soon.
So when should we use these indicators to position portfolios defensively? Historically, excluding the 1973 period, the market has peaked at an average RRI reading of 76 and about four months prior to recession. Of course, there is variability around that average, so as with the RRI and recessions, we can never point to a specific RRI level and say this is when we should “get defensive.” In practice, portfolios should be adjusted slowly over time to reflect the growing likelihood of a recession as the RRI, along with other components of our business cycle analysis and overall investment process, becomes increasingly worrisome.
RRI and Business Cycle Analysis in Context
We believe it’s important to view the RRI in relation to other important economic and market indicators, the combination of which is likely to achieve superior results versus any indicator on its own. This additional analysis adds depth when attempting to understand how markets behave in different economic phases.
Our analysis of shorter-term cycles uses a smoothed series of the Institute for Supply Management (ISM) Manufacturing PMI®, which divides the economic cycle into four phases:
- Recovery: smoothed PMI less than 50 and rising;
- Accelerating expansion: smoothed PMI greater than 50 and rising;
- Slowing expansion: smoothed PMI greater than 50 and falling; and
- Contraction: smoothed PMI less than 50 and falling.
We focus on the manufacturing PMI for a number of reasons, primarily because of its leading properties, highly cyclical nature, large multiplier effect on the rest of the economy, and high correlation with shortterm financial market performance and earnings.
Historically, multiple short-term cycles have occurred within each longer-term cycle. The economic expansion beginning in 2009 offers a good example, with three separate periods of accelerating expansion and three slowing expansionary periods without falling into recession. In this context, the RRI can help determine whether a short-term cyclical slowdown will lead to a more prolonged recessionary period. Since 1970, the PMI indicator has never moved into contraction while the RRI was below 50. During the expansion that began in 2009, each prior cyclical slowdown was accompanied by an RRI far below 50, which correctly forecasted a higher likelihood of reacceleration. As PMIs decline from high levels in a slowing expansion, an RRI above 50 suggests a higher likelihood of economic contraction, in our view.
We use a Hodrick-Prescott (HP) filter to smooth the PMI time series. The HP filter is a mathematical tool used in macroeconomics, especially in business cycle theory, to obtain a smoothed time series, one that is more sensitive to long-term than to short-term fluctuations. Due to the nature of the smoothing calculation, the current value of smoothed data points can change as new data points are added and the series is resmoothed. As such, one should never use the smoothed data series to determine “where we are today” in the cycle. This is better achieved through a more qualitative assessment of the raw PMI data and other macroeconomic observations, including takeaways from our RRI analysis. The primary benefit of the HP filter is to achieve a smoothed data series to categorize historical cycle phases while removing the monthly “noise” of the PMI data series.
|Average Length (Months)||9.0
|Median Lengths (Months)||9.7||19.2||17.2||13.2|
|Average RRI Level||67.2||38.9||47.7||75.9|
|Median RRI Level
|Minimum RRI Level
|Maximum RRI Level||82.1||61.0||86.1||89.0|
Consistent with the expectation that market downturns coincide with economic contractions, the annualized average return for the S&P 500 during a contractionary phase, as defined by the PMI series, has been -1.8% since 1970. This compares to an annualized average return of close to 15% in accelerating expansions. As such, it is critical to determine whether a slowing expansion will reaccelerate or devolve into contraction. An RRI reading far above 50 indicates a high likelihood of the PMI series falling below 50.
From an asset class and sector perspective, we believe the combination of our short-term and long-term business cycle analysis can help in understanding market behavior. As the probability of an economic contraction rises (RRI above 50 and a PMI declining toward the 50 level), it may be advantageous to position portfolios more defensively. In the contractionary phase, for example, the Bloomberg Barclays U.S. Treasury Index has consistently outperformed higher risk asset classes. By comparison, equities have historically suffered losses in that phase, supporting our belief in the value of fully diversified portfolios.
However, not all stocks behave the same during periods of market stress. Analyzing the last three contractionary phases, relative performance across sectors has historically displayed a distinct pattern in times of market stress.
By definition, cyclical sectors, for example, Financials, Consumer Discretionary, and Materials, exhibit high correlations with economic growth and therefore have underperformed during contractionary phases of the business cycle. Conversely, defensive sectors such as Consumer Staples, Utilities, and Health Care have consistently outperformed the S&P 500 over these periods. The contractionary phase associated with the dot-com bubble is an outlier in this regard since both cyclical and defensive sectors outperformed the overall market. In our view, this was a unique period given the S&P 500 drawdown was almost entirely driven by one sector; in fact, the median sector performance over this contractionary period was 5.7% while the Information Technology sector was down 52.8%. Furthermore, given the strong relative performance of cyclical sectors, it indicates to us that many investors were not yet pricing in concerns of an economic contraction; per NBER the official recession did not begin until March 2001.
|1990||2000 (Dot-Com Bubble)
*On September 28, 2018, S&P changed its GICS structure replacing the Telecommunication Services sector with the new Communication Services sector.
This risk-on, risk-off framework also applies when analyzing the relative performance (versus the S&P 500) of previous, significant equity drawdown periods. Given the data, investors are likely to come to the quick conclusion that allocating to more defensive sectors or high quality fixed income would outperform in periods of market stress. While this conclusion is true, it is important to keep in mind many of these drawdown periods came with little warning. Furthermore, as we detailed earlier, drawdown periods which do not coincide with recessions (+/-12 months of drawdown) are normally short lived and viewed as buying opportunities, or at least not the time to panic and reduce equity exposure. However, periods associated with recessions lead to significant performance divergences across sectors and asset classes that may persist for longer periods. While infrequent, insight leading into these periods would allow many investors to exploit these persistent performance divergences and limit the downside effects of a recessionary-led bear markets, as they did in 1990, 2000, and 2008.
This leads us back to the goal of the RRI, which we constructed to assist in evaluating the likelihood of economic recessions as we move through the later stages of the business cycle. As a critical component of our business cycle analysis, the RRI allows us to take a measured approach when formulating future market expectations and make better decisions in periods of market turbulence. We aim to forecast recessionary bear markets since drawdowns outside of recessions are likely to be shallower and shorter in duration. We will follow up with periodic updates on our RRI work detailing what our indicators are telling us about the broader business cycle and our expectations of future recessions.