# CE essentials proxy by income quintile (1984–2023)

Supports Chapter 5, section "A Check Against National Data" (table and text).

**Claim.** Between 1984 and 2023 the food share of total spending fell in every income quintile of U.S. households, while a broader essentials proxy (food + shelter + utilities + transportation + health care) rose in every quintile: about +6.3 points in the lowest fifth, +3.3 in the middle fifth, +0.4 in the highest fifth. Across all households it moved from 62.7% to 64.0%.

**Source.** U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys, annual mean expenditures (all consumer units and income-before-taxes quintiles), via the public BLS API v1. Series ids are listed in `fetch_and_compute.py`. `ce_raw.csv` holds the values exactly as returned.

**Reproduce.**

```bash
python3 fetch_and_compute.py            # fetches only missing series (API v1: 25 queries per day), then computes
python3 fetch_and_compute.py --offline  # recompute from ce_raw.csv
```

Output: `ce_shares.csv` (shares of total expenditure by group and year).

**Limits (also stated in the book).**
- The proxy omits childcare and communications and counts all vehicle costs; it is not the GEC itself.
- It cannot separate price increases from larger homes or better medical care.
- Quintiles are drawn afresh each year; they do not follow the same households, and the lowest quintile includes many retired people.
- Shares are of total expenditure, not income. Reported income in the CE is unreliable at the bottom (spending was 2.1 times after-tax income in the lowest quintile in 2023), so income-based ratios are not used.