What does the divergence in US soft and hard data mean?
Explore the growing divergence between US soft and hard economic data and its implications for consumption, manufacturing, and inflation.
The divergence between soft (survey-based) and hard (directly observed) economic data in the US has complicated the analysis of the economy and asset prices. Analyzing three macroeconomic variables—consumption, manufacturing production, and inflation—puts the current gap in context and paints a potential path forward.
Consumption: US consumers more anxious
Soft data comprises the future expectations of two key consumer surveys: the University of Michigan’s Surveys of Consumers and the Conference Board’s Consumer Confidence Survey®. In contrast, hard data is the real retail sales ex automobiles and fuel, as well as month-on-month personal consumption expenditure. Figure 1 presents both composites, given the inherent volatility in the hard data we also include a smoothed version of it.
Figure 1: US consumption: Soft vs hard data
The difference between the z-score of soft and hard data composites reveals that the soft data has deteriorated relatively more, reaching levels last observed during the Global Financial Crisis (GFC) (Figure 2).
It’s also important to note the divergences between the surveys (Figures 3) and within the surveys (Figures 4 and 5). The headline index levels of the two consumer surveys have historically behaved differently over the business cycle (Figure 3). Generally, the University of Michigan survey has been much more sensitive to inflation, whereas the Conference Board survey has reflected labor market dynamics. The latest University of Michigan survey indicates “it’s never been this bad,” whereas the Conference Board survey suggests “it’s not nearly as good as it was, but it’s still better than average.”
Each survey provides detailed breakdowns into consumer perceptions of their current situation and future expectations (Figures 4 and 5). Comparing these subcomponents shows a gap emerging between future expectations and current perceptions, indicating that consumers expect a deteriorating outlook.
US manufacturing data: Businesses remain positive
Examining US manufacturing production against Manufacturing PMI New Orders (Figures 6 and 7) and Manufacturing PMI’s Output Index (Figures 8 and 9) shows minimal divergence between the hard and soft data. This suggests business expectations remain more grounded than those of consumers.
This reflects the reality that businesses are more active economic agents via capex and hiring decisions, whereas consumers have much less control over their income stream and are primarily controlling only their own spending.
Figure 6: US manufacturing: Soft (new orders) vs hard data
Figure 8: US manufacturing: Soft (output) vs hard data
US inflation: Rising divergence driven by new factors
While the divergence of US soft vs hard inflation data is rising in year-on-year percentage terms (Figure 10), it hasn’t yet reached the peak it hit during the GFC (Figure 11). But there is an important difference in the drivers between the GFC and today. During the GFC, inflation expectations remained static, but actual inflation fell. Today, inflation expectations are rising while hard data is trending lower.
Figure 10: US inflation: Hard data vs soft data
Of course, inflation expectations have limited predictive power, often exceeding realized inflation by 2 to 4 percentage points throughout much of the post-GFC period (Figure 11). Moreover, US consumer inflation expectations historically have been heavily influenced by gasoline prices (Figure 12).
Putting US divergence in a global context
Do other regions show similar divergence in soft and hard data? We performed a similar analysis of macroeconomic variables for Germany and the UK, summarizing results in Figure 13.
Figure 13: Divergence in key macroeconomic variables for the US, Germany, and UK
| Divergence | Macroeconomic Variable | US | Germany | UK |
|---|---|---|---|---|
| Divergence metric quantifies the how far the current reading deviates from the historical norm, measured in standard deviations from a mean of zero* | Consumption | -2.1 | -1.4 | -0.5 |
| Manufacturing production | 0.1 to -0.2 | 0.7 | -1.0 to -1.5 | |
| % YoY soft minus hard | Inflation (%) | 3 | 1.1 | 0.7 |
Source: Macrobond, Bloomberg Finance, L.P., State Street Investment Management, as of April 30, 2025. See appendix for data used. *Z-scores of soft and hard data composites. These composites are constructed by averaging the z-scores of their respective sub-indices, using the common historical sample.
Notably, divergence has emerged in UK manufacturing production, but that is likely due to hard data being bolstered by frontloading before Liberation Day, as confirmed by the Q1 GDP figures in which exports did particularly well.
History’s not repeating, but might it rhyme?
Today’s divergence between US soft and hard data—particularly consumption and inflation—is unusual relative to history and compared to other countries.
Importantly, the deterioration in consumer sentiment hasn’t been matched by a similar reaction on the business side. At least not so far. Since businesses determine employment, and that in turn influences labor income and hence consumption, business expectations can be considered to be more self-fulfilling and matter more for forecasting the future direction of US growth.
Further, inflation divergence is less concerning because current drivers are different from that which caused the high reached during the GFC. Again, today’s rising inflation expectations against a disinflationary trend observed in the hard data contrast with the GFC’s falling inflation and static inflation expectations.
Finally, historically, the returns of major asset classes have not responded to differences in levels of hard vs soft data.
Returns of the S&P 500 Index, Bloomberg US Aggregate Bond Index, and Bloomberg Commodity Index show no meaningful correlation, either positive or negative, with the level of data divergence. Figure 14 summarizes the key statistics of a regression of the monthly returns with the divergence z-scores of the consumption and manufacturing production, and difference in year-on-year percentage inflation.
Figure 14: Impact of divergence in soft vs hard data on major asset class returns
| US consumption divergence in standard deviations | Statistical measures | S&P 500 | US Agg | Commodities |
|---|---|---|---|---|
| Correlation | 0.01 | 0.08 | 0.02 | |
| R squared | 0.00% | 0.60% | 0.00% | |
| P value | 0.91 | 0.13 | 0.72 | |
| US manufacturing production (new orders) divergence in standard deviations | Statistical measures | S&P 500 | US Agg | Commodities |
| Correlation | 0.08 | -0.1 | 0.08 | |
| R squared | 0.60% | 1.10% | 0.60% | |
| P value | 0.25 | 0.12 | 0.26 | |
| US manufacturing production (output) divergence in standard of deviations | Statistical measures | S&P 500 | US Agg | Commodities |
| Correlation | 0.06 | -0.09 | 0.02 | |
| R squared | 0.30% | 0.70% | 0.10% | |
| P value | 0.39 | 0.21 | 0.73 | |
| US inflation divergence in percent | Statistical measures | S&P 500 | US Agg | Commodities |
| Correlation | 0.09 | 0.1 | -0.02 | |
| R squared | 0.70% | 1.00% | 0.10% | |
| P value | 0.08 | 0.05 | 0.62 |
Source: Macrobond, Bloomberg Finance, L.P., State Street Investment Management. Considered gross total monthly returns (unhedged) in USD for S&P 500 index, Bloomberg US Aggregate Index and Bloomberg Commodity Index. For US Consumption: monthly data from February 1992 until April 2025; soft data consists of Conference Board Consumer Confidence Index Expectations and University of Michigan Consumer Expectations: Hard data consists of Real Adjusted Retail Sales Less Autos and Gas Stations and US Personal Consumption Expenditures Chained Dollars. For US manufacturing production: monthly data from May 2007 until April 2025; hard data consists of Manufacturing Production; soft data consists of Manufacturing PMI New Orders Index, and separately, Manufacturing PMI Output Index. For US inflation: Monthly data from March 1991 until April 2025; soft data consists of Conference Board Consumer Confidence Inflation Rate Expectation 12-month and New York Fed Consumer Inflation Expectations Median 1-year Ahead Expected Inflation Rate; hard data consists of US CPI: Urban Consumers. Divergence in standard deviations refers to z-scores of soft and smoothed hard data composites; these composites are constructed by averaging the z-scores of their respective sub-indices, using the common historical sample. Divergence in percent is the difference in soft minus hard data.
The Pearson correlations and R square of the regression equation are all near zero, and the p-value for the slope coefficient is significant only for the Bloomberg US aggregate bond index to US inflation divergence. But even there, the R square is sufficiently small to indicate low predictive power.
These findings are not completely unexpected given that the theory predicts that economic surprises tend to impact equity and bond returns, as opposed to data divergence levels.
So, while the divergence between soft and hard data will continue to grab headlines and fan uncertainty, we believe it's not a major worry or sign of an imminent recession.