Paper short abstract:

We hypothesize that public pressure motivates governments to devalue a currency. We use structural topic models to quantitatively analyze daily Japanese news articles, and employ a supervised machine learning approach to predict foreign exchange market intervention by the Japanese central bank.

Paper long abstract:

When and why does the Japanese government intervene in the currency market? Like other countries in East Asia, the Japanese Ministry of Finance frequently tries to influence exchange rates, ostensibly to limit excess volatility, but in practice usually to depreciate the currency. Technically, this implies selling the Yen and buying a reserve currency such as the USD, accumulating foreign reserves in the process. The decision to intervene is highly political: depreciating a currency helps exporters but hurts importers, and benefits those with investment abroad at the expense of those with savings at home. Yet little is known about what influences the decision to intervene beyond the relative level of the Japanese yen: decision-making is opaque and lobbying, if it takes place, is largely unobservable.

In this paper, we use the public discussion of exchange rate and competitiveness-related topics in the major Japanese financial newspaper (the Nihon Keizai Shimbun) to predict forex intervention. We employ a structural topic model (STM) to quantitatively process the content several hundred thousand news articles, and use a supervised machine learning approach to develop and estimate a model of the probability of an intervention. We find strong evidence that public pressure to intervene precedes action, suggesting that organized economic interest groups are the primary beneficiaries and the likely supporters of foreign exchange intervention.