Research
Work In Progress
Exchange Rate Narratives with Cormun, V.
Abstract: Leveraging Wall Street Journal news, recent developments in textual analysis, and generative AI, we estimate a narrative decomposition of the dollar exchange rate. Our findings shed light on the connection between economic fundamentals and the exchange rate, as well as its absence. Starting from the late 1970s, we identify six non-overlapping narratives explaining exchange rate changes. These narratives include U.S. policy and technological changes in the first half of the sample, and financial market news in the second half.
Working Papers
What Should the Inflation Target Be? Views from 600 Economists with Ambrocio, G., Ferrero, A. and Jokivuolle, E.
CEPR Discussion Paper Series, DP17289.
Media: Suerf policy brief, VoxEU column on the article, VoxEU column on the survey, e-axes article
Abstract: In a survey of more than 600 economists, most respondents prefer their central bank to have an explicit inflation target. Roughly half want the central bank to keep its current target. Two thirds of the rest want to raise the target, with a median preferred change of one percentage point. In a hypothetical scenario in which the central bank has no prior history of inflation targeting, an additional 12% of the respondents would prefer a different (typically higher) target than the current one. This result suggests that the costs of changing the current target hold some respondents back from wanting an actual target change. Respondents who are worried about the central bank credibility are less likely to support a target raise. Conversely, preference for a target raise is more likely to come from those who are concerned about the zero lower bound on the nominal interest rate. The average estimate of the equilibrium real interest rate in the sample is 0.6%. However, personal views about the equilibrium real interest rate do not predict a preference for a target raise.
Bank of Finland Research Discussion Papers 11/2018.
Abstract: The recent financial crises have brought into focus questions regarding the quality of banks' assets. We study the patterns in banks reserving for and reporting of loan losses in the EU before and after implementation of the Single Supervisory Mechanism (SSM). We find that banks that 1) have less tier 1 capital, 2) are smaller, 3) are less liquid and 4) have smaller net interest margins either report relatively smaller loan loss reserves or less loan losses, even after including various controls. This supports the hypothesis that financially weaker banks may have a larger incentive to engage in balance sheet window dressing. We further find that the SSM has reduced but not eliminated the under-reserving and under-reporting bias. In addition, there has been a separate positive effect on the overall proportion of nonperforming loans (NPLs) that are realised as losses among the banks that have been under direct supervision by the SSM since implementation of the SSM.
Publications
Narrative Triggers of Information Sensitivity
Quantitative Finance, 2024.
Presentations: RiskLab/BoF/ESRB Conference on Systemic Risk Analytics 6/2023, EEA-ESEM 2023
Abstract: Economic research has shown that debt markets have an information sensitivity property that allows these markets to work properly when price discovery is absent and opaqueness is maintained. Dang, Gorton and Holmström (2015) argue that sufficiently "bad news" can switch debt to become information sensitive and start a financial crisis. We identify narrative triggers in the news by utilizing machine learning methods and daily information about firm default probability, the public's information acquisition and newspaper articles. We find state-specific generalizable triggers whose effect is determined by the language used by journalists. This language is associated with different psychological thinking processes.
A Thousand Words Tell More Than Just Numbers: Financial Crises and Historical Headlines with Roukka, T. and Nyberg, H.
Journal of Financial Stability, February 2024
Presentations: TSE Economics Seminar 9/2021, BoF Financial Stability Department Seminar 2/2022, RiskLab/BoF/ESRB Conference on Systemic Risk Analytics 5/2022, EEA-ESEM 2022 , 4th conference on Non-traditional Data, Machine Learning, and Natural Language Processing in Macroeconomics 2022 , ASSA 2023
Data: Historical economic news article topic series
Abstract: We show that financial crises are preceded by changes in specific types of narrative information contained in newspaper article titles. Our novel international dataset and the resulting empirical evidence are gathered by integrating information from a large panel of economic news articles in global newspapers between the years 1870 and 2016 with conventional macroeconomic and financial indicators. We find that the predictive information of newspaper article titles that signals coming crisis episodes is substantial over and above the macroeconomic and financial indicators. Feature contribution analysis and crisis case studies reveal that the new indicators capture more detailed, but still generalizable information on the buildup of crises.
Are bank capital requirements optimally set? Evidence from researchers’ views with Ambrocio, G., Hasan, I. and Jokivuolle, E.
Journal of Financial Stability 50, 2020.
Media: SUERF policy note, Official page for the survey, VoxEU blog post on the survey Preliminary results, VoxEU blog post on the preliminary results
Abstract: We survey 149 leading academic researchers on bank capital regulation. The median (average) respondent prefers a 10% (15%) minimum non-risk-weighted equity-to-assets ratio, which is considerably higher than the current requirement. North Americans prefer a significantly higher equity-to-assets ratio than Europeans. We find substantial support for the new forms of regulation introduced in Basel III, such as liquidity requirements. Views are most dispersed regarding the use of hybrid assets and bail-inable debt in capital regulation. 70% of experts would support an additional market-based capital requirement. When investigating factors driving capital requirement preferences, we find that the typical expert believes a five percentage points increase in capital requirements would “probably decrease” both the likelihood and social cost of a crisis with “minimal to no change” to loan volumes and economic activity. The best predictor of capital requirement preference is how strongly an expert believes that higher capital requirements would increase the cost of bank lending.
Scandinavian Journal of Economics 120(1) 2018, 31-62.
Abstract: Studies of the early warning systems (EWSs) for banking crises usually rely on linear classifiers, estimated with international datasets. I construct an EWS based on an artificial neural network (ANN) model, and I also account for regional heterogeneity in order to improve the generalization ability of EWS models. All of the banking crises in my test set are then predictable at a 24-month horizon, using information from earlier crises. For some countries, estimation with a regional dataset significantly improves the predictions. The ANN outperforms the usual logit regression, assessed by the area under the receiver operating characteristics curve.
Journal of Banking and Financial Economics 1(5) 2016, 121–143.
Abstract: CDS spreads are often used as markets view of credit risk. There is no popular alternative to it; perhaps only the distance-to-default measure based on Merton (1974) comes close to it. In this paper we investigate the relationship between these two measures for large European banks in post subprime crises era. The analysis makes use of conventional Granger causality test statistics for individual banks and for the whole panel data. As for the results, we find that the lead-lag relationship between these variables varies over time and over different banks and economic regimes. The lead of distance-to-default is stronger for banks in problem countries (PIGS), during European debt crises, for relatively small banks and when there are large changes in CDS spread. These results suggest that we may have predictive power by not only using the CDS spread, but also other measures such as the distance-to-default.