EPU Indices

Categorical EPU

US Categorical EPU Indices

Trade Policy Uncertainty

Monetary Policy Uncertainty

Korea Categorical EPU

Greece Categorical EPU

Japan Categorical EPU

Climate Policy Uncertainty

US-China Tension Index

Other Indices

Twitter-based Uncertainty

US Equity Market Volatility

Infectious Disease EMV

Financial Stress Index

World Uncertainty Index

Sustainability Uncertainty

Financial Stress Indicator

Firm-Level Political Risk

Immigration-Related Indices

Brazil Economic Uncertainty

Geopolitical Risk Index

Inter-Korean Geopolitical Risk

Economic Uncertainty Queries

Firm-Level Instruments

Energy Uncertainty Indexes

Google Uncertainty for India

Oil Price Uncertainty Index

Cable News-based EPU

Pakistan Climate Uncertainty

Climate Risk Indexes

Financial Stress Index

Download Data                                            Download Paper

We are pleased to host the Financial Stress Index (FSI) by Hites Ahir, Giovanni Dell'Ariccia, Davide Furceri, Chris Papageorgiou, and Hanbo Qi. They construct the FSI for 110 countries for the period 1967 to 2024.

Conceptually, they follow Bernanke (1983) and Romer and Romer (2017) and aim at classifying as episodes of financial stress in which an economy experiences an increase in the cost of credit intermediation or disruptions to the credit supply. As described by Romer and Romer (2017), the rise in cost of credit intermediation includes both a higher cost of funds for financial institutions relative to a safe interest rate and an increase in other operational costs associated with their lending activities. Put differently, they identify episodes in which, for a given level of the expected return on safe assets, the cost (quantity) of credit to the economy increases (decreases). Note that this definition excludes reductions in the supply of credit stemming from increases in interest rates due "normal" cyclical factors such as tighter monetary policy.

They follow a three-step process to construct the index. First, similar to Romer and Romer (2017), they try to narrow down the amount of text in the Economist Intelligence Unit (EIU) country reports to search for information about financial stress. As Romer and Romer (2017) states: "To narrow the amount of the volumes we need to study closely, we start with a keyword search for terms likely to appear in accounts of financial distress. The most important are bank and financial, but we also search for crisis, rescue, bailout, crunch, and squeeze." Romer and Romer (2017) also mention that they experimented with searching for "credit".

They adopt a similar approach and identify paragraphs/lines containing two set of keywords: (i) credit, financial, bank, lending, and fund, and; (ii) crisis, crunch, squeeze, bailout, rescue, tight, contract, and reluctant. The words in the first group (credit, financial, bank, lending, fund) aims at capturing discussion related to the "financial" market, while the words in the second group tries to capture the distress part (e.g., rise in the cost). This approach is very important to limit the volume of pages they need to read for each country-period pair, since they do this process for 110 countries, four times a year (quarterly frequency), and the average length of the report per country is about 30 pages. So, they cover a very high amount of text compared to Romer and Romer (2017).

In the second step, they read the paragraphs extracted in step 1 to confirm that the text is indeed describing developments associated with contemporaneous financial stress. The point here is to exclude false positives.

In the last step, they sum the verified signals of financial stress in each period to convert the qualitative classifications to quantitative measure of financial stress. Their working assumption is that as the severity of financial stress increases so will the extent to which it is covered in the EIU country reports. So that higher word counts will correspond to more severe episodes. While this assumption could be questioned, its main advantage is that it improves the transparency and replicability of their approach and reduces the risks of using previous knowledge to focus on certain periods or to quantify the level of financial stress based on ex-post information. Their also later verify their assumption by comparing our index to existing financial stress indicators such as Romer and Romer (2017) that rely on expert judgement to assess the severity of a stress episode.

An obvious difficulty with these raw counts is that the overall length of country reports varies across time, and across countries. Thus, to make the index comparable across countries, they scale the raw counts by the total number of words in each report and rescaled by multiplying by 1,000. Two factors further help improve the comparability of the index across countries. First, the index is based on a single source. Second, the reports follow a standardized process and structure. In addition, the process to put together the reports described earlier helps to mitigate concerns about the accuracy, ideological bias and consistency of the index.