Consumer and business sentiment affect everything from momentum in stock markets, to elections, to purchasing decisions. But what factors drive consumer and business sentiment? To answer that question, we looked at measures of sentiment — also known as confidence — and their underlying determinants going back to the 1980s. We found that the factors that have historically accurately signaled the direction of sentiment are no longer reliable.
We examined the University of Michigan Consumer Sentiment Index (UMCSENT), the Consumer Confidence Index (CCI), and the Business Confidence Index (BCI). We then pulled data on various macro factors. These included unemployment, interest rates (Fed funds rate), inflation, GDP growth, loan delinquency rates, personal savings rates, stock market returns, and labor force participation rates.
Next, we regressed each of our consumer and business sentiment measures against each of the macro variables, partitioning the sample by decade. Figure 1 presents the results for our model using UMCSENT as the dependent variable. Figure 2 uses CCI, and Figure 3 uses BCI. In the tables, a “+” symbol denotes that the coefficient in our model was significant and in the correct direction, (i.e., based on historical expectations). An “x” symbol denotes that the coefficient was either insignificant or in the incorrect direction (i.e., not what we have seen historically).
Figure 1. University of Michigan Consumer Sentiment Index (UNCSENT)
Figure 2. Consumer Confidence Index (CCI)
Figure 3. Business Confidence Index (BCI)
The first interesting finding is that in our consumer sentiment measures during the 1980s, almost all the variables were significant and in the direction you would expect. GDP growth led to great consumer confidence; greater unemployment led to lower consumer confidence; greater inflation led to less consumer confidence, etc. But as time went on, our model became less predictive. By the post-COVID period, an increase in GDP did not lead to an increase in consumer sentiment. An increase in unemployment also had no impact on sentiment. In fact, only two variables out of eight had significant power in predicting the direction of consumer sentiment: inflation and the stock market returns.
To put some numbers to the coefficients in our model, during the 1980s a one percentage point increase in inflation led to a 3.4-point drop in the Michigan index, and a 1% increase in unemployment led to a 3.6 drop in the Michigan index.
Indeed, during the post-COVID period our model has become much more muted. From 2020 forward, a 1 percentage point increase in inflation led to just a 1.1-point drop in the Michigan index, and a 1% increase in unemployment led to just a 2.3 drop in the index.
Further, the strength of our model (i.e. the predictive power) has also decreased over time. The Adjusted-R^2 was 0.88 in the 1980s and dropped to 0.72 in the present day. We see similar results in the BCI model as well but not to the same degree that we see in our consumer sentiment results.
What may be the underlying cause of all this? There are likely many factors, but one highlighted by past literature could be partisanship. Individuals have noted that individuals switch their views on the economy and sentiment to a much greater extent in the present day based on who holds political office. The upcoming US presidential election could be one of the underlying factors that we omitted in our study.
Whatever the case, unemployment, labor force participation, and GDP growth no longer explain how consumers are feeling about their prospects. The root causes of this phenomenon deserve more careful study.
