Examining Market Volatility Around Economic Data Releases and Recessions
Introduction
Market volatility dynamics have persistently perplexed investors over the course of the current economic cycle. Disruptions stemming from the Global Covid Pandemic, followed by the first serious bout of inflation that the US economy has experienced in 40 years, and the subsequent remarkably vibrant but uneven economic recovery, have all conspired to repeatedly defy markets consensus. Against this backdrop, the attention being paid to macro-economic variables has surged, but the thinking surrounding the impact of these factors on volatility markets is muddled. This paper seeks to systematically examine the behavior of volatility in response to key economic data releases, as well as to investigate how that behavior changes heading into recession. Our analysis has two main takeaways. The first key conclusion is that economic releases tend to reduce, rather than increase, implied volatility. The second is that volatility only sees a modest increase as the economy enters into a recession. We note that volatility measures only really pick up later in the economic adjustment process, well after the economic contraction begins.
Gauging Market Volatility
There are many ways to measure volatility, but for the purposes of this paper, we focus on two key metrics when discussing market volatility for equities and rates: the VIX index and the MOVE index, respectively. Using these measures as proxies has several advantages, most importantly their widespread familiarity and their long histories. Studying market variables over economic cycles requires data that goes back more than one cycle of expansion and contraction. Both the VIX and the MOVE indices have data going back to (at least) January 1990, the point at which our analysis starts.
To measure the impact of discrete economic data announcements, this paper focuses on seven key macro-economic reports. Parsing through the types of economic data releases, we focus on those with history going back to the start of 1990, are closely followed by market participants, and have release dates that must be known in advance. Based on this selection criteria, our analysis uses the following releases: Retail Sales, Manufacturing ISM, Services ISM, Non-Farm Payrolls (NFP), Consumer Price Index (CPI), Industrial Production (IP), and Personal Consumption Expenditure (PCE) (Figure 1).
Figure 1: Economic Data Indicators, Release Dates, and Sources
We also examine the period leading into a recession and the period shortly after the start of the recession. The dates used for the start of a recession are the official designations published by the National Bureau of Economic Research (NBER). For this analysis, we looked at the three months prior to the start of a given recession and following six months. Going back to January 1990, our sample period includes four recessions, but this study excludes the Covid contraction because it was a purely exogenous shock induced by the pandemic. This paper is more interested in endogenously-generated economic recessions, since they are far more typical and to some degree more predictable. The data period ends in June 2024, therefore covering a total of 34.5 years (Figure 2).
Figure 2: Volatility and Recessions
Setting the Volatility Record Straight
Economic data is generally volatility-suppressing. Market participants often warn of the possibility that some forthcoming piece of economic data may surprise markets and therefore catalyze a surge higher in volatility metrics. But over time, we find that this is a losing bet. Both the VIX and MOVE indices consistently see more declines than increases on days when key economic data is released (Figure 3).
This is the case for most releases – not just as a straight tabulation of up days versus down days, but even in comparison to the average trading day. Volatility behavior is not normally distributed, meaning volatility indices typically see more down days than up days. Over our sample period, the VIX and MOVE indices ended the day higher roughly 45% of the time (dashed line in Figure 3). With the exceptions of both the Manufacturing and Services ISMs as well as PCE, the volatility indices see more frequent declines on days when the data is released relative to the moves on a typical day.
Figure 3: Percentage of Volatility-Up Days on Data Release Days
A reasonable counterargument to this line of thinking is that simply looking at up-versus-down days is too simplistic, since magnitudes matter. Theoretically, fewer but more significant up days could still make it worthwhile to be long volatility heading into big data releases. It turns out however that this is not the case either. Even in magnitude-adjusted terms, we found that most economic data releases are still volatility-suppressing the majority of the time. This is true for both the VIX and the MOVE indices (Figures 4 & 5). The only meaningful exceptions are the Services ISM and PCE for equity volatility, and both the Manufacturing and Services ISMs for rates volatility.
The second major takeaway here is that the NFP release is particularly volatility-suppressing. As evidenced in Figure 2 above, VIX and MOVE indices respectively closed lower 70% and 77% of the time on NFP days. Similarly, the declines in both equities and rates volatility in response to NFP releases were far more significant than the impact from any other economic data, in either direction (as evidenced in Figures 4 & 5). This is especially true on the rates side, in both mean and modal terms. There are not many reliable rules-of-thumb in markets, but selling rates volatility heading into an NFP release seems to come pretty close (Figure 6).
Figure 6: Change in MOVE Index on day of NFP Release
Why might this be the case? Quite simply, markets price in a risk premium ahead of a given NFP announcement out of fear of a possible surprise. If that fear does not materialize, that risk premium gets priced out and volatility drops. Clearly, as seen by the results, more often than not, it does not. This is particularly remarkable because NFP data is notoriously hard to predict, and therefore is a constant source of surprise to investors. This has remained the case even since the onset of the Covid pandemic, when NFP surprises have persistently been larger (Figure 7).
Figure 7: Magnitude of Surprise to NFP (Median Estimate – Realized, in thousands)
Why the ISMs and PCE are partial exceptions to the broader trend is a more complicated question, and a topic that requires a separate deeper dive. One possible hypothesis is that ISM releases are the most timely and forward-looking major economic releases that the market gets. As a result, they carry more new information for investors than the other releases, and perhaps therefore have greater capacity to influence the macro narrative. Though PCE is less timely, it is comprehensive and may therefore also be more impactful in this sense. In any case, more investigation is required here.
Volatility in the Time of Recession
Not all stages of the economic cycle have the same characteristics, and not all volatility spikes are created equal. Another piece of conventional wisdom is that volatility picks up as economic data deteriorates heading into a recession. This is only partially true, and to a lesser degree than is commonly believed. Looking across the last three endogenously driven recessions (1990, 2001, and 2007-2009), volatility measures did increase after the economy entered recession, but the increase was modest. What is more notable, implied volatility saw an initial peak shortly after the official start of a recession, and then steadily declined. Here too, the pattern holds across both equities and rates (Figure 8).
Figure 8: VIX and MOVE indices entering into recession (t = 0 at start of recession)
Interestingly, this is a generalizable finding across different cycles; the same basic pattern holds across all three recessions. In all three cases, the VIX and MOVE indices were making new lows about 90 calendar days after the start of recession, both dropping below where they were in the month leading up to said recession (Figure 9 & 10). This was then followed by a period of resurgent volatility in the subsequent 90 days after the start of a recession, though the magnitude of that surge varied considerably. Why this is the case may be due to idiosyncratic factors unique to each recession, but the main takeaway here is that investors do not need to try and time the start of recession to be long volatility. That opportunity often comes later once the recession has already begun.
Figures 9 & 10: Both VIX (left) and MOVE (right) initially rose modestly at the start of a recession, but then hit local lows
Extending this point further, it is worth highlighting that the particularly large volatility spike during the Global Financial Crisis happened well after the recession had started (Figure 11). This is the surge in volatility that investors often associate with a significant economic contraction and remember years later. Yet that panic point did not occur until October 2008, 10 months after the recession had begun (and past the 6-month time horizon this study looked at, as our goal is to focus on the particular volatility dynamics during the period surrounding the start of recessions).
Figure 11: The surge in volatility during GFC, 10 months after the recession had started
Looking at the volatility response to different economic releases heading into recession also offers some useful lessons. Once again, the NFP release stands out here, but in a different way. Unlike normal markets where NFP is equity volatility-suppressing, it is a clear source of higher volatility heading into recession. In all three recessions in our sample, NFP drove the VIX higher the first month of each recession (Figure 12). In each case, a poor NFP print led investors to downgrade their growth views and price in greater downside tail risks. Notably, the following NFP release led to a drop in VIX in each case.
Figure 12: Change in VIX on NFP days heading into recession (t = 0 first month of recession)
The ISMs and PCE stand out here too as being the only other releases that have been associated with a net higher VIX during downward turns in the cycle (Figure 13). The impacts of the manufacturing ISM and PCE however are modest and uneven. Only the Services ISM really qualifies as a strong signal (Figure 14). The Services ISM has driven VIX higher consistently at the start of recessions, which makes intuitive sense since its release represents the newest information about the largest share of the US economy.
Interestingly, the takeaways are completely different on the rates side. The only meaningful positive contributor to the MOVE index during these periods is the CPI release (Figure 15).
Figure 15: Change in MOVE heading into recession
On net, CPI has driven rates volatility higher heading into and shortly after the start of recession with remarkable consistency (Figure 16). Conversely, NFP was true to form in being a persistent volatility-suppressor (Figure 17). We believe there is an intuitive explanation for this: all else equal, the rates market is more sensitive to inflation dynamics than it is to growth dynamics, whereas the opposite holds true for the equity market.
Conclusion
Having a sample size of just three recessions has obvious drawbacks and limits the generalizability of any conclusions. Nevertheless, this paper highlights two key findings, namely (1) economic releases tend to reduce, rather than increase implied volatility, and (2) volatility only sees a modest increase as the economy is entering recession, only really rising later in the economic adjustment process. It also challenges conventional wisdom on the interaction between volatility and macro-economic data, outlining the logic behind these results. Lastly, we identify areas for further study, including a more rigorous quantitative assessment of the relationship between some of the key macroeconomic indicators and volatility metrics, an extension of this analysis to examine the behavior of different tenors across the term structure, and an expansion of this analysis to other asset classes.