01 / SENTIMENT
No robust sunshine signal.
None of the 12 adjusted country/specification tests passes the 5% false-discovery threshold. This does not establish that the true effect is zero.
Examine the intervals ↗An empirical study · 2015—2024
More sunshine.
More confidence?
I’m interested in psychology and decision economics: how mood and our surroundings shape everyday choices. This project began with a personal question—could I find a trace of that in weather, confidence and spending data?
Erkam Sait Aygün · Sabancı University · DSA210
01 / The sentiment recordOECD · consumer confidence
A decade, in perspective.
GermanySpainTurkeyUK
The person behind the question
It started with
personal curiosity.
Psychology · decision economics · data science
I study data science with a strong interest in psychology and decision economics—the way mood, environment and macroeconomic pressure quietly shape the everyday choices people make.
I wanted to move past intuition and look at whether those forces actually leave a measurable trace in real data. That curiosity is why I built this project.
480country-month observations
4 / 3countries with CCI / retail data
6future-year validation windows
12robustness specifications
What the evidence says
The interesting result
is the limit.
The analysis connects my initial curiosity with the evidence. Associations, uncertainty and model comparisons each tell a different part of the story.
02 / PREDICTION
Last month sets a high bar.
CCI persistence achieves RMSE 0.49. The full weather–macro forest does not outperform it in calendar-based validation.
Compare the models ↗03 / SPENDING
Weather is not the whole story.
The combined retail forest has held-out R² 0.42. A matched macro-only model performs almost identically.
Inspect the ablation ↗A transparent research record
Follow every number
back to its source.
Inspect country coverage, download the research panel and out-of-fold predictions, or read the methods and Turkish research report.