Between March 2009 and September 2018, the S&P 500 Index returned 333%, marking the longest-ever ‘Bull-Market’ on record. Market prices then retreated 19.8% in the following 11-weeks only to gain back more than half of the difference in 2019. Since World War II, there have been 13 ‘Bear-Market’ stretches, where U.S. stock prices fell in excess of 20% from a prior high-point, and seven have resulted in economic recessions. In the other six instances, the U.S. economy did not experience an economic contraction in the following two years. That means we have arrived at a cross-road.
The difference between the two divergent paths is the presence of job growth. The U.S. economy has only contracted, when job losses have caused an increase in the unemployment rate. This relationship is evident in the last two Bear-Markets, when Bull-Market peaks fell within 113-days of the first-reported contraction in nonfarm payroll data that was not immediately revised upwards by the U.S. Bureau of Labor Statistics (BLS) and followed by a positive reading in the next month. Given the outsized influence of Quant Funds, we expect this historical precedent to repeat itself, when the current 101-month streak of consecutive job gains, the longest on record, comes to end. What’s more, there is reason to believe that that endpoint will arrive during the next three months; a concern because the prolonged period of U.S. fiscal stimulus has driven outsized increases in asset prices that are susceptible to a sharp downward reversal.
To understand our minority view, it is helpful to know that, by gradually edging-up interest rates eight times since December 2016, the U.S. Federal Reserve Open Market Committee (FOMC) is hoping to engineer what economists call a ‘soft-landing,’ where the goal is to lift rates, such that asset prices and employment are slowed-down without triggering an economic recession. In August 2018, FOMC Chairman, Jerome Powell, recognized this delicate dance, saying: “The FOMC has been navigating between the shoals of overheating and premature tightening with only a hazy view of what seem to be shifting navigational guides.” If his words suggest to you a risky and low-probability outcome, then you are correct. The FOMC has only achieved a soft-landing on one occasion, in the 1994/1995 period, when the unemployment rate was deemed to be above the ‘natural rate’ of sustainable employment.[1] By comparison, the FOMC estimates the current unemployment rate (of 3.8%) to be 0.7 percentage points (pp) below the natural rate of sustainable employment, which means an economic slow-down can only be realized, in the face of higher unemployment. This reality poses a problem to the FOMC, because the U.S. economy, “has always ended up in a full-blown recession,” whenever the unemployment rate has increased by 0.3pp, according to William Dudley, former New York Federal Reserve president. As a result, the FOMC is attempting something that has never been done before.
When nonfarm payrolls are concerned, what’s important here is that a slow-down in credit has, historically, translated into a guaranteed set of layoffs in the construction sector, which is highly-sensitive to rising interest rates. To borrow from Goldman Sachs: “Far from being a surprise, a meaningful slowdown in residential investment is a key feature of a successful tightening.”[2] The impact on construction layoffs is further well-understood. This point was emphasized in 2014 by former FOMC chairman, Alan Greenspan, who told a panel that construction has historically represented 8% of the U.S. workforce, adding: “You knock out half of that, that’s 4 percentage points. That’s all the increase, major increase in the unemployment rate.”[3] Because of the roughly two-year timeframe to complete a building project, there is also an inherent lag-effect between rate increases and layoffs in the construction sector, where, “between 500,000 and a million construction workers [flow] into unemployment, on net, in each 12-month period, as part of the natural dynamics of the labor market,” according to a Federal Reserve analysis of the last recession noted.[4] After all, today’s building projects are funded with ‘yesterday’s loans’ in the construction sector, where workers roll-off completed projects into the unemployment lines, when loan growth slows. This lag-effect formed by the building cycle explains why economic recessions always ‘seem’ to materialize out of thin-air. On November 14, Chairman Powell acknowledged this lag effect, which he estimated at 12-18 months, saying: “When we make a decision, and we raise interest rates, it actually takes some time for the full effects of that change to show up in the real economy.” Investment in the U.S. housing sector has fallen in four consecutive quarters, so the process in already in motion.
With higher interest rates putting downward pressure on job growth, Wall Street research analysts are broadly expecting nonfarm payroll gains to decelerate through the 2020 period, including Goldman Sachs that is forecasting payrolls to decline from the 3-mo. average (of 186k/mo.) to 110k/mo. in December 2019 and 65k/mo. in June 2020 and Bank of America that is forecasting average payroll growth of 176k/mo. in 2019 and 110k/mo. in 2020. By comparison, we expect the growth in nonfarm payrolls to fall-off much more sharply, in a manner that is more consistent with prior contractions. The difference of opinion is simple. Wall Street research forecasts rely on regression models that cannot, by the admission of their authors, account for ‘structural changes’ to employment and ‘exogenous shocks’ that are outside the scope of historical pattern data. These factors, which are detailed below, are also critical to understanding the forward picture.
From our vantage, the biggest risk to nonfarm payroll growth is the structural shift to seasonal employment in the retail sector. Per Exhibit 1, U.S. employers added 714k seasonal workers in 2018, a 16% increase (or 100k) from the 614k hired in the 2017 calendar year, according to data from Challenger, Gray & Christmas, Inc. (CGC). These seasonal hires represented 71% of the 1mm announced hires in 2018, a sharp rise from 2017, when the same figure was just 56%, so the job mix has been shifting towards part-time hires. What’s more, seasonal hires in 2018 were highly-concentrated in the top-10 employers that, together, account for 643k or 90% of the total, marking a 38% increase (of 178k) from 2017. The same employers have well-defined seasonal programs of 3-4 months that began in October/November, implying a new trend line, in response to the increased need for online fulfillment services. The absence of these hiring schemes is the reason why we anticipated the rapid contraction in job growth during the month of February, when BLS reported nonfarm payroll gains of just 20k/mo., a figure that was well below the Wall Street consensus forecast of 180k/mo. The trend also makes a short-term rebound unlikely.
Exhibit 1. Seasonal Employees
Source: Challenger, Gray & Christmas, December 6, 2018.
Against the above, it is important to recognize here that the shift to seasonal employment has fundamentally altered data in ways that may not be fully-appreciated by Wall Street analysts. Seasonal jobs not only give the impression of favorable nonfarm payroll gains, but also distort jobless claims data because part-time workers are not entitled to receive unemployment benefits in most U.S. states. As a result, the traditional relationship between declining nonfarm payrolls and increasing jobless claims has broken-down. This partly explains why the Wall Street consensus missed the nonfarm payroll figure in February (of 20k/mo.) by 160k/mo. February data also shows the number of people ‘employed part-time for economic reasons’ declining by 837k/mo. (to 4.3 million), after rising by 500k/mo. in January, which BLS said, “may have resulted from the partial federal government shutdown.” But, this account (at best) only explains a piece of the data. After all, the decline in part-time workers in February (of 837k/mo.) exceeds the 800k furloughed government workers, while it seems unlikely that 63% of said government workers (of 500k) gained a new job, after January 11, when their paychecks temporarily ceased.
In addition, we suspect the influx of seasonal workers has clouded the labor force participation rate, which BLS revised upwards in January to show a 0.2pp increase in the second-half of 2018, offsetting an increase of equal-magnitude in the unemployment rate that rose from 3.7% to 3.9% in the same period. The revised BLS data, which implies the number of people outside the labor force fell by 1 million in six-months, the largest such decline on record, not only dovetails with the 714k seasonal hires in 2018, but also falls outside the scope of U.S. policymakers. Chairman Powell highlighted this point earlier this month, when he said the recent increase in the labor force participation was “an upside surprise that most people didn’t see coming.” Taken together, these data points suggest the labor force participation rate is susceptible to unfavorable revisions in the coming months, which will have a meaningful impact on payroll data.
Exhibit 2. Announced Job Data
Source: Challenger, Gray & Christmas, March 7, 2019.
Without seasonal hires boosting top-line growth, we can expect nonfarm payrolls to turn negative during the next several months, if there is a pick-up in corporate restructurings and bankruptcies, which are considered to be ‘exogenous shocks’ by Wall Street economists. This notion is consistent with prior contractions in the labor market. Economists at Goldman Sachs note: “Historical evidence suggests that absent an outside shock, the market is unlikely to decelerate slowly on its own.”[5] To understand the impact, one needs only to look at nonfarm payroll date from July 2018, when the bankruptcy of Toys ‘R’ Us triggered 32k layoffs and monthly nonfarm payroll data (of 157k/mo.) was 31k below the Wall Street consensus estimate of 188k/mo. Per Exhibit 2, CGC data shows announced job cuts of 130k in the first two-months of 2019, marking a 62% increase (from 80k) during the same period in 2018. These figures indicate an acceleration in layoffs from 2018, when CGC reported total job cuts of 539k, up 29% from the 419k identified in 2017. What’s more, announced job cuts far outweigh new hires (of 89k), so there is a contraction underway that did not exist in 2018, when new hires totaled 182k in the period. These job losses also extend beyond the distressed retail sector into staple product categories, including names like Verizon (10k layoffs) and General Motors (14k layoffs). Given the industry supply/demand imbalance, additional layoffs should be expected in the automotive sector, which has added 350k payrolls since the cycle-bottomed at 970k employees, and the U.S. oil sector, where companies have rehired many of the 160k employees that were let go in 2014, when oil prices last swung below the point-of-profitability for U.S. energy producers.[6] Meanwhile, at least 20k (of ~63k) employees at Sears, Inc. are likely to be let go in the next two-months.
Given the above trendlines, we think it is reasonable to assume the combination of (i) fewer seasonal hires and (ii) accelerating layoffs from ‘one-time’ corporate events will, together, cause nonfarm payrolls to contract during the next several months. This understanding is juxtaposed with the Wall Street consensus that cite the record number of U.S. job openings (see Exhibit 3) as evidence that the U.S. labor market will continue to accelerate during the next two-years. However, the same trend was observable in the financial downturn of 2000, when the unemployment rate was last pushed-down below the 4.0% level. The conflicting data, in our view, can be explained by the fact that most unfilled jobs are located in areas with high property values. This notion is observable today in Silicon Valley, where many employers are building new regional offices in areas that have more affordable housing options than California.[7] Academic research further supports the claim, including research from Peter Ganong and Daniel Shoag, who have found that, “for low skill workers, rising house prices have eroded the gains from migration.”[8] The Joint Center for Housing Studies of Harvard University also reported that residential mobility reached an all-time low in 2018. Meanwhile, recent data from the Federal Reserve shows unemployment in rural areas to be far higher than in urban areas, where levels are now below pre-crises levels. As such, in rural areas, only 78% of 25 to 54-year-olds are employed, which compares to over 90% in urban areas.
Exhibit 3. U.S. Unemployed vs. Unfilled Positions
Source: U.S. Bureau of Labor Statistics.
The forward trend outlined above is a concern, because nonfarm payroll data is widely-understood to be a strong forward indicator of recession, so a negative print has the ability to cause the kind of widescale reassessment among investors that has, historically, accompanied a market tipping-point. This relationship was evident in January 2008, when nonfarm payrolls contracted precisely one-month after the official end of the business cycle and 11-months before the National Bureau of Economic Research (NBER) publicly reported the recession. The link is also intuitive. Federal Reserve data shows 40% of Americans cannot cover a $400 emergency expense, so a large portion of the labor force is ‘living paycheck-to-paycheck.’ And, when there are fewer workers receiving checks from employers, the immediate response is declining consumer spend (equal to 2/3 of the U.S. economy) and missed payments on home and auto debt obligations. To that end, U.S. holiday sales in December fell -1.2% from the prior month, the largest such decline since 2009, while New York Federal Reserve data shows that in 2018 auto delinquencies (90 days late) reached 7 million, the highest on record.
An analysis of the initial BLS data releases (or ‘preliminary results’) indicates the predictive power of nonfarm payrolls extends beyond recessions to Bear-Markets, a more meaningful data point to investors, who prioritize avoiding equity losses over identifying economic contractions. After all, had one correctly identified the recession that began in March 2001, you still would have suffered losses exceeding 20%, because the Bear-Market began a year earlier. Per Exhibit 4, the S&P 500 Index reached its Bull-Market zenith within 113 days of the first BLS report containing negative nonfarm data that was not immediately reversed and followed by a positive monthly print. For example the S&P 500 Index peaked on March 24, 2000 and (105 days later) on July 7, BLS reported job gains of +206k/mo. for June alongside a downward revision for May that lowered monthly payrolls gains from +231k/mo. to -165k/mo., alongside a corresponding increase in the unemployment rate that rose from 3.9% to 4.0%. During the last downturn, the S&P 500 Index reached its high-point on October 11, 2007 and (113 days later) BLS reported, on February 1, 2008, a contraction in payrolls (of -17k/mo.) for January. This figure was revised upward in the following month, but only after the downward trend was confirmed by a February payroll figure of -163k/mo.
Exhibit 4. Monthly Nonfarm Payroll Data (final prints)
Despite the significance between negative payrolls and Bear-Markets, this relationship has been overlooked by Wall Street analysts, who are likely performing analyses based on the finalized figures from BLS, which revises its data three-times, on account of the 100k/mo. margin of error. Instead, the link can only be found in the ‘preliminary’ figures that represent the ‘real-time’ information influencing market participants, rendering it more powerful, in our view. Whatever the exact nature of their connection, the data shows that the S&P 500 Index is unlikely to record a new high for several years, after a contraction in job growth. It’s also clear the contractions in payrolls appear suddenly and the reversals are large.
We stress the relationship between nonfarm payrolls and Bear-Markets here because the outsized presence of automated trading strategies suggests the next downturn will be data-driven. This implies a tighter relationship between negative payrolls and equity market declines. Ranging from simple passive strategies to complex hedge fund products, Morgan Stanley data shows Quant Funds commanding assets under management of (at least) $1.5 trillion, while J.P. Morgan estimates 85% of total trading volumes are related to automated trades from Quant Funds and Index Funds. The share of daily trading volumes attributable to Quant Funds (see Exhibit 5) has also steadily increased from 14% in 2010 to 29% in 2018, when volumes eclipsed retail investors, according to the Tabb Group. This means we are dealing with a rewired marketplace, where the principal actors are algorithms, which have been programmed to ‘sell’ at the first-available moment of a poor data release. This is not a random accident, but a response to the “data dependent” monetary approach adopted by former FOMC chairman, Janet Yellen, so investors have been trained to focus on income data.
Ultimately, Quant Funds pay sizable fees to stock exchanges to gain a split-second lead over retail investors, and we expect them to use that advantage, when a negative nonfarm payroll is released, which makes an outsized reaction all but certain. To give an idea, just consider the events of last February 5, when an unexpected increase in wage growth data (of 2.9%) reported by BLS caused the VIX index, which measures the 30-day implied volatility on the S&P 500, to spike 100% in a single day, marking its largest movement since the Wall Street Crash of 1987. The VIX spike caused total losses in three exchange-traded note funds and a 10% equity correction in the first-half of 2018, so the events of that day were far from minor.[9] When payrolls contract, the market response will be far worse.
Taken together, investors should be prepared for the Bull-Market to close-out during the next several months alongside a negative nonfarm payroll report. The market response to the event is likely to be swift, given the automated strategies of Quant Funds and the misplaced optimism in the U.S. labor market among retail investors, who will pile on to any sell-off With household equity allocations and price/earnings valuation multiples (see Exhibits 6-7) at levels only observed during the 2000 Technology Bubble, the downward price action is also likely to be severe, such that losses will be frontloaded, particularly if the trend is confirmed by a negative print in the following month.
Let us remember here the words of famed investor Bernard Baruch, who bluntly said: “I made my money by selling too soon.” We recommend a version of that strategy today.
Exhibit 5. Daily Trading Volumes (2010-2018)
Source: Tabb Group dated December 2018.
Exhibit 6. Financial Asset Allocation Among Investors by Quarter (1996-2018)
Source: Goldman Sachs Research, January 2019.
Exhibit 7. S&P 500 Index Risen 6x Faster than U.S. GDP
Source: Based on data from FactSet, S&P Global, and U.S. Federal Reserve
[1] In response to fears over rising inflation that has risen to 2.5%, the FOMC, under Chairman Alan Greenspan, doubled short-term interest rates (to 6%) without triggering a recession.
[2] Hatzius, Jan et al. “The Housing Slowdown.” Goldman Sachs Economics Research dated November 4, 2018.
[3] “A Conversation with Greenspan.” Council on Foreign Relations, October 29, 2014.
[4] Paciorek, Andrew. “Where are the Construction Workers?” Federal Reserve Board, February 26, 2016.
[5] Mericle, David. “Engineering A Soft Landing.” Goldman Sachs Economics Reserved, April 15, 2018.
[6] North American automakers are operating at an 82% capacity utilization rate, according to LMC Automotive, which implies supply capacity exceeds demand by 3.2 million vehicles, while U.S. oil rigs have fallen by 7% (from 885 to 824) in the last three-months, according to Baker Hughes data.
[7] In California, the median house price in 2018 was $570k, more than double the national average, according to the California Association of Realtors, which says the median home price in the Bay Area is above $1 million. More than half of renters and over a third of mortgaged homeowners in California spend in excess of 30% of their income on housing, per data from the state housing department.
[8] Ganong, Peter and Daniel Shoag, “Why has Regional Income Convergence in the U.S. Declined?” NBER Working Paper 23609, July 2017.
[9] The VelocityShares Daily Inverse VIX Short-term ETN was liquidated by its sponsor, Credit Suisse, on February 5, when its NAV moved 80% in a single session, causing total investor losses of $2 billion.