DSA210 / Evidence

Sunshine and consumer confidence.

The primary specification compares sunshine with consumer confidence, within each country, after accounting for month-of-year and a linear trend.

No causal claim

Weather observations, aggregated cities and revised economic series support an observational study. HAC uncertainty and multiple-testing adjustment improve inference; they do not identify a causal effect.

Country-specific adjusted sunshine coefficients and 95 percent HAC confidence intervals — Estimated CCI-index-point difference per additional 100 sunshine hours, controlling for calendar month and linear trend. HAC covariance with 12 lags; intervals are pointwise 95%, not simultaneous. Download SVG ↗

The primary estimates

Country	N	CCI points / 100h	95% lower	95% upper	HAC p	BH q (12 tests)
Germany	120	-0.750	-1.723	0.222	0.131	0.313
Spain	120	0.655	-0.696	2.005	0.342	0.456
Turkey	120	-0.121	-2.411	2.169	0.918	0.918
UK	120	-1.494	-3.349	0.361	0.114	0.313

All q-values exceed 0.05. A non-significant result is not proof of no effect; the intervals show the uncertainty compatible with this design.

Geography and trend sensitivity

One adjustment family: 4 countries × 3 predeclared specifications = 12 tests.
Country	Specification	Effect / 100h	HAC p	BH q
Germany	Weighted cities	-0.750	0.131	0.313
Germany	Capital city	-0.445	0.308	0.456
Germany	First differences	-0.040	0.186	0.371
Spain	Weighted cities	0.655	0.342	0.456
Spain	Capital city	0.948	0.088	0.313
Spain	First differences	0.237	0.074	0.313
Turkey	Weighted cities	-0.121	0.918	0.918
Turkey	Capital city	0.873	0.285	0.456
Turkey	First differences	-0.082	0.680	0.785
UK	Weighted cities	-1.494	0.114	0.313
UK	Capital city	-1.569	0.103	0.313
UK	First differences	-0.024	0.720	0.785

Weighted cities vs. capitals

The main weather measure averages five cities per country with fixed population weights. The capital-city series provides an alternative geographic measure. Divergent estimates warn against treating a country as a single weather station.

First differences

The sensitivity specification relates month-to-month changes, with seasonal controls and HAC(12) covariance. It reduces level/trend confounding but does not resolve omitted variables or establish causality.