One of the current conundrums is why overall assessments of the economy have not risen in line with actual developments (for example, as summarized by the Misery Index), and why these advances have not rewarded the incumbent’s approval ratings (discussed here). I don’t have an answer, but I have some observations.
Economic conditions and (measured) economic sentiment
First, how far do the current readings of the Michigan Economic Sentiment Index (FRED series UMCSENT) deviate from expectations over the past 30 years? Figure 1 shows the actual UMCSENT series (red), fitted series (green) and residuals (blue).
figure 1: Actual values, fitted values, and residuals from UMCSENT regression.
returns (in my previous article Regarding this question) is (gasoline price is deflated by CPI):
UMCSENT = 44.26 – 4.06suffering – 16.4pgasoline
adjust-R2 = 0.61, SER = 8.30, DW = 0.28, Nobs = 379, sample 1992M01-2023.07. bold italic Indicates significance at 10% MSL, using HAC robust standard errors.
(I use the Bloomberg consensus for the July unemployment rate, and the Cleveland Fed’s 8/1 July CPI flash forecast.)
Carola Binder Excellent discussion of sentiment, current and expected partisan influence. Here, I focus on the differential impact of the misery index components (unemployment rate and year-on-year CPI inflation rate) on the disaggregated sentiment index. First, here are the different series of pictures.
figure 2: University of Michigan Economic Sentiment Index (bold black), Democrats (blue), Independents (yellow green), Republicans (red). Recession peak-to-trough dates as defined by NBER are grayed out. Source: University of Michigan, via FRED, University of MichiganNational Bureau of Economic Research.
Note that Republicans’ views were much more positive than those of Republicans under Trump, but have turned much more negative under Biden. This reflects the well-known trend that sentiment is much higher when the president is a member of the respondent’s political party than otherwise. It would be interesting to see if the effect is stronger in recent times than in previous decades, but there is no fairly continuous time series prior to 2006, so it is difficult to answer that question.
As of July, the Republicans’ approval rating was 51, while the Democrats’ approval rating was 92.
How do Democrats, independents, and Republicans view the state of the economy differently based on perceptions of unemployment and inflation? I investigate this by regressing various UMCSENT indicators on these variables, as well as a dummy variable for the Democratic president.
UMCSENTIt = one0 + a1ut + a2PIt + a3backt
Where I = Overall, Democrats, Independents, Republicans, 2006M01-2023M07. UMCSENT divided by 100, ut and PIt are in decimal form. Table 1 lists the results.
Note: Bold indicates significance at 10% msl using HAC robust standard errors. The F-test (p-value) is a constraint that the unemployment rate and the inflation coefficient be the same p-value.
Column 1 shows the UMCSENT (overall) results from FRED. The coefficients on unemployment and inflation are about the same size. The coefficient is negative for Democratic presidential administrations: A Democratic president is associated with a 0.05 reduction in UMCSENT (standardized index mean of 0.86). The results are roughly the same if the sample is restricted to the same period in which we had categorical results (column 2); therefore, the period in which we have categorical data does not appear to be an anomaly.
Democrats appear to be placing slightly less emphasis on inflation than on unemployment, but statistically, not significantly less. In contrast, Republicans place significantly more weight on inflation than on unemployment (the F-test nearly rejects zero values for the coefficient of equality).
The coefficients on the dummy variables corroborate the finding that when the presidency is held by someone from the same party/faction, the perception of the state of the economy is more positive: the coefficient for DEMPRES is positive (+0.191) for Democrats and positive for Democrats (+0.191). Negative for MPs (-0.361) and slightly less negative for Ind (indicating that Ind is more like a MP than a Democrat). In general, standardized coefficients (“beta” coefficients) indicate that the largest effects are associated with dummy variables.
There is an asymmetry in the influence of presidential partisanship. During this period (2006 to 2013), Republicans’ presidential affiliation had a huge effect—0.361 vs. 0.191 (in absolute terms).Other things being equal, they real When a Democrat is president, they will be negative about the situation.
Interestingly, while both Indian and Republican sentiment depend on unemployment and inflation (Adj R2 0.5 or so), Democratic sentiment is largely unexplained (Adj R2 about 0.04). The actual year-over-year growth rate of the S&P 500 appears to have a much greater impact on the Democrats (as measured by the “beta” coefficient) than on the Reps or Indiana.
(Measuring) Economic Confidence and Biden’s Approval Ratings
It is widely believed that Biden has not been praised for the strong performance of the economy. This is verified by looking at Biden’s approval trend (from FiveThirtyEight) and the sentiment index.
image 3: University of Michigan Economic Confidence Index (bold black, left scale), Democrats (blue, left scale), independents (yellow green, left scale), Republicans (red, left scale), Biden support trend, % (light green, right scale). Source: University of Michigan, via FRED, University of Michiganfive thirty-eight.
Using the relationship between UMCSENT and Biden approval ratings for the period November 2021 to July 2023, Biden approval ratings in July should be 43.1, not the 538 trend value of 40.5.
