# Gauge 1 · The Lipstick Test

**Published research summary · November 5, 2026.** Research status: completed historical analysis and working paper. Observation window: January 1959–August 2026. The working paper has not undergone journal peer review.

When budgets tighten, do small luxuries hold up better than larger discretionary purchases? The Lipstick Test examines that idea through US consumption history. Its main comparison is between BEA’s combined cosmetics, perfumes, bath, and nail category and clothing and footwear. The data cannot isolate perfume, lipstick, luxury brands, or individual shoppers.

The study follows nine business-cycle peaks since 1959. For each category, it measures growth twelve months after the peak relative to the path implied by its preceding growth, then compares the two categories. A positive gap means the grouped cosmetics category performed better relative to its own prior path. The twelve-month endpoint can fall during recovery as well as recession. Analysis choices were locked in a local Git commit before confirmatory calculations; this was a local analysis plan rather than an external registry.

The primary average gap is **+1.788 log points**, with an adjusted placebo p-value of **0.06519** and a rough 95% bootstrap interval of **−1.859 to +6.717**. Only four of nine gaps are positive. **The registered primary test does not establish the small-luxury hypothesis at the 5% threshold.** The result is sensitive to the earliest episode: removing 1960 changes the mean to −0.342 log points.

Two other registered comparisons are more favorable. Cosmetics growth is less sensitive to aggregate consumption growth than clothing growth in the quarterly regression: slope difference −1.109, 95% HAC interval −1.801 to −0.417, adjusted p=0.00254. The cosmetics-minus-jewelry post-peak mean is +5.725 log points, adjusted placebo p=0.02819, although its rough episode bootstrap interval includes zero. These answer different questions and do not replace the primary clothing comparison. None identifies household substitution or a causal psychological mechanism.

The accompanying descriptive indicator is **+0.446 percentage points in August 2026**, classified as steady. It measures the difference in year-over-year quantity growth; it is distinct from the trend-adjusted event statistic above. The category’s breadth, nine historically linked episodes, the short pre-history for 1960, calendar-placebo assumptions, and future national-accounts revisions limit interpretation. The evidence supports a qualified historical comparison, with the inconclusive primary finding kept visible.

## Chart choices

1. **Lead:** `charts/gauge01/g01_scoreboard.svg` — show every recession gap so the uneven results are visible.
2. `charts/gauge01/g01_betas.svg` — explain the distinct pooled quarterly association.
3. `charts/gauge01/g01_tilt.svg` — use the continuous indicator with its August 2026 date and percentage-point units.

## Source and verification note

BEA NIPA underlying-detail quantity, price, and expenditure tables; FRED/BLS price and labor series; NBER chronology. Frozen raw files have October 2, 2026 UTC pull dates, corresponding to October 1 acquisition in New York; BEA tables identify a September 30 revision. Cached replay and independent numerical checks are recorded in `reports/gauge01-verification.md`. The completed manuscript is `papers/gauge01-lipstick/manuscript.md`; the public working paper is [Cosmetics and Recession Resilience](https://ryderxue.com/research/cosmetics-recession-resilience/paper.html). Preserve all three registered tests and reported robustness in the linked technical material.
