A bank distrust index
No index to start from
The model needed distrust in banks, country by country, and no dataset had it. Heyert and Weill (2024) had used Google Trends, which ranks searches inside each country and cannot compare two countries. I kept their idea and turned it into a count.
- 13terms of banking stress, from bank run to bailout: the five of Heyert and Weill, widened
- 106languages spoken in the sample, every term translated by hand
- 954keywords to search, once the duplicates were removed
- 150countries queried one by one in Google Ads Keyword Planner
- ÷internet users (World Bank): yearly searches per 1,000 internet users
Fig. 1. From an idea to one number per country.
Source: thesis, section 3.2.1, and the indicator’s workbook · Method: Keyword Planner, broad match, average monthly searches from January 2021 to January 2022, times 12, per 1,000 internet users (World Bank population in 2021 times the percentage using the internet). The thesis says per 100,000, but the workbook multiplies by the percentage as a number from 0 to 100, so the unit is per 1,000 · Limit: search habits, Google’s share of searches and censorship differ across countries: the index measures searching, not distrust itself.
Keyword Planner says it counts searches in every language. A test said otherwise: the same term gave different volumes in English and in the local language. So every term went into every language of the sample.
- Bank RunEnglish
- Corsa BancariaItalian
- Ruée BancaireFrench
- BankansturmGerman
- Corrida BancariaSpanish
- Corrida BancáriaPortuguese
- 挤兑Mandarin Chinese
- 取り付け騒ぎJapanese
- 은행 인출 사태Korean
- اندفاع السحبArabic
- ריצה על הבנקHebrew
- Đổ Xô Rút TiềnVietnamese
Fig. 2. One of the 13 terms, bank run, in 12 of the 106 languages.
Source: the indicator’s translation sheet · Method: dictionaries and terminology databases (PanLex, UNTERM, Glosbe, Tatoeba, Google Translate) · Limit: some terms have no real equivalent in every language. Where none existed, the English was kept and filtered out later.
Where it landed
- Bank concentration34.0%26.8%
- Collectivism index12.9%11.7%
- Political instability14.6%11.0%
- Bank distrust (this index)5.5%5.5%
- Population growth0.1%5.0%
- Access to electricity3.3%4.6%
Fig. 3. The six largest shares of the thesis model’s explained variance (R² = 0.503). The new index is fourth.
Source: thesis, Table 3 · Method: LMG and PMVD shares of R², 10,000 sampled orderings, 144 countries · Limit: a share of explained variance is not an effect. The index is fourth by share, but its coefficient is not significant.
- 4th of 29regressors by share of explained variance: 5.5% under LMG
- p-value = 0.26coefficient negative and not significant, the opposite of the expected sign
More distrust, less adoption: the thesis reads it as distrust keeping people away from decentralised finance too, and leaves it as a hypothesis, because the coefficient is not significant.