Unemployment Expectations as a Leading Indicator

Executive Summary

This analysis examines whether US consumer expectations about future US unemployment, as measured by the University of Michigan Survey of Consumer Attitudes (SCA), serve as a leading indicator for actual changes in the unemployment rate. We test the hypothesis that expectations about “more unemployment” reported 6 months earlier are correlated with current unemployment rates, suggesting that consumer sentiment may anticipate labor market conditions.

Key Finding: The analysis reveals a statistically significant correlation of 0.257 between expectations of more unemployment (lagged 6 months) and current unemployment rates, based on 207 monthly observations from July 2008 to September 2025.

Current Data Snapshot

Latest Readings (as of October 2025): - Unemployment Rate: NA% - Expectations of More Unemployment: 64%

Recent Data: Past 13 Months

Unemployment Rate and Expectations: Past 13 Months
Month Unemployment Rate (%) Expect More (%)
October 2024 4.1 32
November 2024 4.2 32
December 2024 4.1 40
January 2025 4.0 48
February 2025 4.1 51
March 2025 4.2 66
April 2025 4.2 65
May 2025 4.2 64
June 2025 4.1 57
July 2025 4.2 57
August 2025 4.3 63
September 2025 4.4 65
October 2025 NA 64

Data Sources

  • Unemployment Rate (UNRATE): Monthly unemployment rate from the U.S. Bureau of Labor Statistics, accessed via FRED (Federal Reserve Economic Data)
  • Consumer Expectations: University of Michigan Survey of Consumer Attitudes, Table 30 - “Expected Change in Unemployment During the Next Year”
    • The “More” variable represents the percentage of respondents who expect unemployment to increase in the next year

Methodology

The Leading Relationship Concept

The core question we address is: Do consumer expectations about unemployment predict future unemployment changes?

To test this, we employ a lagged correlation approach:

  1. Expectations Variable: We use the percentage of consumers expecting “more unemployment” from the Michigan SCA survey
  2. Outcome Variable: We use the actual unemployment rate (UNRATE) from BLS
  3. Temporal Alignment: We lag the expectations variable by 6 months, meaning:
    • Expectations from month t are compared to unemployment in month t+6
    • This tests whether expectations 6 months ago are associated with current unemployment

Why 6 Months?

The 6-month lag is chosen to align with the survey question’s time horizon (“during the next year”) and to account for the typical lag between economic expectations and realized outcomes. This period allows for: - Information processing and decision-making delays - Hiring/firing cycles in the labor market - The time it takes for economic conditions to materialize

Statistical Approach

We calculate: - Pearson correlation coefficient between lagged expectations and current unemployment - Statistical significance using a two-tailed correlation test - 95% confidence intervals for the correlation coefficient

Results

Correlation Analysis

Correlation Analysis: Expectations (6-month lag) vs Current Unemployment
Correlation P-value CI Lower CI Upper N Observations Start Date End Date Significant?
0.257 0.000188 0.125 0.38 207 2008-07-01 2025-09-01 Yes

Interpretation: - A positive correlation of 0.257 indicates that when expectations of more unemployment were higher 6 months ago, current unemployment tends to be higher. - The correlation is statistically significant (p = 0.000188), suggesting a meaningful relationship.

Scatter Plot: Expectations vs. Unemployment

This scatter plot shows the relationship between expectations of more unemployment (6 months earlier) and current unemployment rates. Each point represents one month. The blue line shows the linear trend, with the shaded area representing the 95% confidence interval.

Key Observations: - Weak positive relationship between the two variables - The relationship is statistically significant, suggesting that consumer expectations do contain predictive information about future unemployment

Individual Series: Unemployment Rate

Individual Series: Expectations of More Unemployment

Time Series: Standardized Comparison

This time series plot shows both variables standardized (expressed as standard deviations from their respective means) to facilitate comparison. The expectations series is lagged by 6 months to align with the unemployment outcomes.

Key Observations: - Both series show similar patterns over time, with peaks and troughs that tend to align - The alignment suggests that consumer expectations may indeed anticipate changes in unemployment - Major economic events (recessions, recoveries) are visible in both series

Discussion

What This Means

The statistically significant correlation between lagged expectations and current unemployment suggests that:

  1. Consumer Expectations as Information: Consumer expectations about unemployment contain information about future labor market conditions. This aligns with the idea that consumers, as economic agents, may have access to information or intuition about economic trends.

  2. Leading Indicator Potential: The 6-month lag structure suggests that expectations may serve as a leading indicator—information available today (expectations) helps predict conditions 6 months in the future (unemployment).

  3. Self-Fulfilling Prophecies: There may be a self-reinforcing mechanism where expectations influence behavior (e.g., reduced spending, delayed hiring), which in turn affects actual unemployment outcomes.

Limitations

Several important limitations should be considered:

  1. Correlation vs. Causation: A correlation does not establish causation. The relationship could be driven by:

    • Common underlying factors (e.g., economic news, policy changes)
    • Reverse causality (though the lag structure mitigates this)
    • Omitted variables

  2. Time Horizon: The 6-month lag is somewhat arbitrary. Different lags might yield different results, and the optimal lead time may vary across economic cycles.

  3. Survey Methodology: The Michigan SCA survey reflects consumer sentiment, which may be influenced by media coverage, political factors, and other non-economic variables.

  4. Sample Period: The analysis covers a specific time period. The relationship may vary across different economic regimes (e.g., expansion vs. recession).

Policy and Practical Implications

If consumer expectations do predict unemployment changes, this has implications for:

  • Economic Forecasting: Expectations data could be incorporated into unemployment forecasting models
  • Policy Making: Early signals from expectations could inform policy responses
  • Business Planning: Firms might use expectations data to anticipate labor market conditions

Conclusion

This analysis provides evidence that consumer expectations about unemployment, as measured by the University of Michigan SCA survey, are associated with future unemployment rates. The positive correlation of 0.257 suggests that when consumers expect more unemployment, actual unemployment tends to increase in subsequent months.

However, correlation does not imply causation, and further analysis—including causal inference methods, robustness checks with different lags, and examination of the relationship across different economic regimes—would strengthen these findings.

Technical Details

Data Processing

All data processing and analysis were performed using R. The analysis script (expectation_lead_unempl.R) performs the following steps:

  1. Loads UNRATE data from FRED
  2. Loads Michigan SCA Table 30 data
  3. Merges datasets by date
  4. Creates 6-month lagged expectations variable
  5. Calculates correlation and statistical tests
  6. Generates visualizations

Output Files

All intermediate and final outputs are saved to the outputs/ directory:

  • merged_unrate_sca_data.csv: Full merged dataset
  • sca_only_data.csv: Filtered dataset with SCA observations only
  • sca_ft_time_series.csv: Time series with YoY changes and lags
  • sca_ft_standardized.csv: Standardized time series for visualization
  • correlation_dataset.csv: Dataset used for correlation analysis
  • correlation_summary.csv: Summary statistics
  • scatter_plot_expectations_vs_unemployment.png: Scatter plot visualization
  • time_series_standardized.png: Time series comparison

Reproducibility

To reproduce this analysis:

  1. Ensure raw data files are in data/raw/:

    • UNRATE_data.csv
    • sca-table30-on-2025-Nov-24.csv

  2. Run the analysis script:

    source("scripts/expectation_lead_unempl.R")

  3. Compile this report:

    rmarkdown::render("report_sources/unemployment_expectations_analysis.Rmd")

Sources: University of Michigan Survey of Consumer Attitudes, U.S. Bureau of Labor Statistics (via FRED)

Analysis Date: 2025-11-25