The Federal Reserve Bank of New York's Economic Heterogeneity Indicators have revealed a growing divergence in consumer spending patterns between college-educated and non-college-educated American households, a trend the researchers characterize as consistent with a "K-shaped economy."
The analysis draws on data from Numerator, an analytics firm that maintains an elective panel of approximately 200,000 U.S. consumers. The panel is reweighted to match U.S. Census Bureau demographics across multiple dimensions — income, education, race/ethnicity, age, and urban status — and the researchers note that their spending metrics align with Census Bureau aggregate retail trade surveys.
The central finding: by December 2025, college-educated households demonstrated nominal retail spending growth approximately 2.4 percentage points faster than non-college households since January 2023. The authors note that "most of the divergence between the two sets of households" took place by spring 2024, suggesting the gap emerged relatively early and has since stabilized rather than continuously widening.
When adjusted for inflation using demographic-specific price indices — a methodologically significant choice — non-college households experienced approximately 4% real spending growth, while college-educated households saw approximately 6%. The use of demographic-specific indices matters because different groups face different inflation rates. Lower-income households spend proportionally more on food and energy, which experienced some of the sharpest price increases during 2023-2024, meaning aggregate inflation measures may understate the cost burden on non-college households.
Several methodological caveats deserve attention. First, the data source is an elective consumer panel — participants opt in, introducing potential self-selection bias. People who join spending-tracking panels may differ systematically from the general population in spending habits, digital engagement, or shopping patterns. While reweighting addresses known demographic imbalances, it cannot correct for unmeasured behavioral differences between panelists and non-panelists.
Second, the study measures retail spending only, excluding housing, healthcare, education, and other major expenditure categories where cost burdens diverge dramatically between education groups. A household spending 40% of income on rent versus 20% faces fundamentally different economic pressures, regardless of retail spending trends.
Third, the "K-shaped economy" framing, while compelling, risks oversimplification. The observed spending divergence could reflect multiple interacting factors: income differences, wealth and asset accumulation, geographic concentration in high-cost versus low-cost areas, industry composition effects (remote-friendly vs. in-person work), differential savings rates, and varying access to credit. The analysis does not disentangle these mechanisms, making the causal story more complex than the headline suggests.