Assuming a normal distribution of surprises, approximately 68% fall within one standard deviation of the mean when standardized using z-scores. See appendix for more color around our use and calculation of z-scores.

The following charts present R Square values for different post-release windows, being the proportion of variance in trading volume explained by economic surprises for FX and equity futures (Figure 1) and options (Figure 2). These data are supplemented with tables detailing coefficients, t Stats, and P-values (see Figure 3 for FX and Figure 4 for equity) for different post-release windows. For instance, our model accounts for 38% (R Square = 0.38) of the variation in FX futures trading volume within the first 5 minutes (8:30:00 - 8:34:59) following 8:30:00am ET data releases (Figure 1).

For the 1-, 5-, and 10-minute intervals, equity options had a higher R Square than FX. However, the R Square for FX options was higher for the day period than equity. This could indicate that trading in equity options is more automated than FX given its faster reaction time to the data.

The intercept is the average daily trading volume on Mondays, assuming no surprises in economic data vs market expectations (i.e., releases matching market expectations), or on days with no releases. Mondays are, on average, the lowest volume days with about 89,333 to 198,182 fewer FX futures trades (Figure 3), and 81,400 to 270,233 fewer equity futures trades (Figure 4) than other days of the week. Wednesdays are typically the busiest day of the week in terms of volumes for FX futures and options. Thursday is typically the busiest day for equity futures and Friday for equity options.

Following the 8:30:00am ET release, surprises in labor, inflation, and retail sales data showed statistically significant impacts on trading volumes in the first 1-, 5-, and 10 minutes. This significance was determined using a P-value threshold of 0.05 (5%). According to our model, there is a 2E-06 (0.000003%) chance that the observed relationships between surprises in Unemployment Rate and equity futures trading volume during the 1-minute interval (8:30:00-8:30:59) was due to random chance (Figure 4).

The 1-minute results were -interesting. On days with no economic data releases or when data showed no surprises versus consensus, in the first minute after the report (8:30:00 - 8:30:59), typically about 1,704 FX futures contracts and 3,395 equity futures were traded (the intercept). A one standard deviation surprise in NFP in either direction led to 1,704 + 18,132 or about 19,836 FX futures (Figure 5), and 3,395 + 14,810 or about equity 18,205 contracts traded between 8:30:00 - 8:30:59 (Figure 6).

In the minute after the reports’ release, surprises in related labor market data, such as the unemployment rate and average hourly earnings, also produced strong impacts on FX and equity volumes as did initial jobless claims ranging from 7,000 to 18,132 in additional FX futures volumes (Figure 5) and 4,000 to 16,000 additional equity futures volumes (Figure 6) for a one standard deviation miss from consensus. 

Like interest rates, retail sales was the second most influential piece of data for FX after the employment numbers A one standard deviation surprise versus forecasts on retail sales typically produced an additional 6,766 contracts of futures volume in the minute after release (Figure 5). 

Interestingly for equities futures, surprises in all the inflation reports produced stronger additional trading volumes in the minute after the release than retail sales. This was especially prevalent for Core CPI, having double the impact of other inflation prints as 9,117 additional contracts traded (Figure 6). Core CPI strips out volatile components of CPI such as food and energy, providing a more measured impact of inflation on consumers. Markets tend to place additional weight on the Core CPI reading than headline CPI due to this. Additionally, the inflation prints relative to employment reports yielded a greater proportional impact in terms of volume traded for equities than interest rates or FX. 

Within 5 and 10 minutes of the release of the various data series, increases in trading volumes tended to produce results of similar statistical significance with the amount of additional trading volume increasing with time. In the 10 minutes between 8:30:00 and 8:39:59 ET, typically there would be 11,030 FX (Figure 7) and 28,671 equity (Figure 8) futures traded. In the 10 minutes after a one standard deviation surprise on employment related data, there were typically anywhere from 18,466 to 49,638 and 10,880 to 66,502more contracts worth of FX (Figure 7) and equity (Figure 8) futures trading volume respectively. Here too, surprises in retail sales were the second largest volume driver for FX and Core CPI for equity, with a one standard deviation surprise adding on average 15,607 to FX futures volumes (Figure 7) and 293 to options volumes (Figure 3); 34,193 to equity futures volumes (Figure 8) and 6,380 to options volumes (Figure 4) respectively.

NFP influenced FX Futures in the 1, 5, 10-minute windows more than the other components of the employment report such as the unemployment rate and average hourly earnings. Initial jobless claims had a more muted impact on equity markets than FX. 

However, in the 1, 5, 10-minute windows, the unemployment rate and average hourly earnings had slightly stronger impacts on equity futures volumes than NFP.

Recognising that FOMC policy announcements can significantly influence trading activity, we included a dummy variable to account for these days. These announcements were statistically significant and yielded strong trading volumes across FX and equity markets. FX and equity futures daily trading volume is, on average, 215,748 (Figure 9) and 990,768 (Figure 10) on FOMC announcement days compared to non-FOMC days. FX and equity options daily trading volume is, on average, 11,515 (Figure 9) and 177,744 (Figure 10) contracts higher on FOMC days.

Source: CME Economic Research Calculations

Source: CME Economic Research Calculations

Source: CME Economic Research Calculations

Source: CME Economic Research Calculations

Since 2021, financial markets have experienced significant uncertainty about future interest rates, mostly because inflation rose sharply after the pandemic. Core PCE, the Fed’s preferred inflation measure, exceeded the 2% target, hitting 3.1% in May 2021 and peaked at 5.3% in March 2022. Core PCE was 2.9% in January 2025.

We calculated z-scores using a three-year rolling standard deviation to normalize the magnitude of surprise relative to historical volatility. For example, to calculate the z-score for the January 2021 data releases, we used the standard deviation from January 2018 to December 2021. This standard deviation is rolling, so the next would use February 2018 to January 2021, and so on. This reflects how the market’s reaction to surprises evolved over time, ensuring we are not using future data to make inferences about past data.

We used dummy variables to account for the differences in trading volumes that structurally occur on different days of the week, using Monday as the baseline. This means Monday is not included in the regression, and when its Tuesday, for example (Tuesday = 1, Wednesday = 0, Thursday = 0, Friday = 0), then the increase or decrease in volume is relative to Monday.

Similarly, we included a dummy variable for FOMC announcement days. This coefficient is the difference in average trading volume on FOMC days compared to non-FOMC days (FOMC day = 1, non-release day = 0).