Blogs review: The Events Study methodology
What’s at stake: The event study approach – a methodology in finance and economics used to detect the presence of event-induced returns within a period – has become ubiquitous in recent debates about the impact of unconventional monetary operations. But its identifying assumptions are generally not very well understood. In this review, we explain the […]
What’s at stake: The event study approach – a methodology in finance and economics used to detect the presence of event-induced returns within a period – has become ubiquitous in recent debates about the impact of unconventional monetary operations. But its identifying assumptions are generally not very well understood. In this review, we explain the different steps followed by researchers to perform event-study analyzes; we point out some of their pitfalls, based on recent discussions following the latest Jackson Hole meeting; and we illustrate its growing popularity in a variety of fields.
The econometric skeleton of an event study
The underlying idea of an event study is to use the high frequency trading values of financial securities to assess the impact of news. If financial markets are efficient, in the sense that information about the future payoffs of the assets are factored in their prices, an event that affect these future payoffs should translate into an immediate repricing. The impact of an event can thus be measured by examining security prices surrounding the event.
Craig MacKinlay writes in a 1997 paper of the Journal of Economic Literature that the event definition represents the first key step for conducing an event study. Starting with the selected event in period t, the interval of time t-1 and t+1 is defined as period over which the event occurs and the asset price will be analyzed. This interval – named “event window” – is usually larger than the specific event of interest because it is aimed at containing the ex post reaction of markets closed in t as well as possible ex ante reactions due to market expectations.
Once the event and the event window are defined, abnormal returns – the difference between the actual ex post return of the asset and the normal return over the event window – are estimated. There are several approaches for estimating normal returns, i.e. the returns that one would have expected in the absence of the event. Norman Strong gives an overview of the different approaches used to estimate the benchmark for abnormal returns. Some assume a very simple data generating process – for example that the ex ante expected return for a security is a constant that can vary across firms – while others use a more complicated structure inspired by economic models like the CAPM or arbitrage pricing theory. After selecting these models, the normal return is estimated over a period of time called “estimation window” and defined in the range t-2 and t-1.
How much of the effect is due to news?
James Hamilton writes – about the impact the quantitative easing – that the hope is that, over the short period studied, the policy announcement itself is the most important news item to which markets were responding. But the problem plaguing any effort to measure the effects of the policy is the fact that the policy was certainly itself also a response to the news of a rapidly weakening economy, which would have been a reason for falling yields even if the Fed had done nothing. In his recent Jackson Hole paper, Michael Woodford points that the fact that the stock market has also declined in value during each of the periods when there were sustained declines in long-term bond yields supports this interpretation.
Michael Woodford furthermore points that the assumptions underlying the event studies methodology is inconsistent with the existence of the policy effects that the study is intended to demonstrate. Taking the sum of the market movements on these announcement days only as a measure of the cumulative effect of the program as a whole — rather than, say, the cumulative change in long-term interest rates over the entire period of the program — depends on believing that the program should only influence bond prices at the times when there is news that changes the expected size of the program, and that the effects of news are (nearly) immediate and permanent. The latter assumptions are familiar ones in event studies in financial economics, of course; but it is important to recall that the justification for this familiar methodology is an assumption that securities markets are “efficient,” so that expected returns looking forward from any point in time are essentially constant (which requires that the effects of news on prices be realized instantaneously and not be subsequently reversed). Such an assumption is not obviously consistent with the existence of effects of Fed purchases on the prices of securities that the study is intended to demonstrate; for if the quantity purchased influences the price of a security, then the market is not efficient in the Samuelson-Fama sense.
Beyond the problem of event clustering – which can be partly dealt by having a small event window – Michael Woodford points out that event studies cannot disentangle the channels through which the event has an effect. The one-day window is narrow enough that, in most cases, one can – for example – plausibly argue that the FOMC’s statement was the only big news affecting fixed-income markets that day; but it is less obvious that the only news in the statement was information about the likely size of the asset-purchase program. In particular, if the statement also contained information that changed expectations about the future path of the federal funds rate, then bond yields should have changed on those days, even in a world where there are no portfolio-balance effects.
Joseph Gagnon writes a seriously wonky rebuttal to Michael Woodford’s skepticism. Some Fed statements and actions clearly had no implications for the path of future short-term rates, and yet they still affected some asset prices. This evidence comes from a range of investigations, not solely from event studies, and the conclusion is not sensitive to changes in the underlying model used to identify movements in the term premium.
A research approach with growing popularity
The events-study methodology was introduced by Ray Ball and Philip Brown (1968) and Eugene Fama et al. (1969) in the late 60s, in the field of corporate finance. Ball and Brown assessed the impact of the information contained in the earnings whereas Fama et al. studied the effects of stock splits. Alongside the analysis of firm stock price’s reaction, a vast branch of the literature implemented event studies for the impact assessments of macroeconomic event or monetary policies decisions on bonds and exchange rate.
Many studies analyzed the impact of credit rating agencies decision. Most of them focus on the price reaction of both corporate and sovereign bond prices on rating decision on sovereign creditworthiness (see Vassalou and Xing for some references). Reisen, H. and J. von Maltzan (1999) examined the announcement of credit rating agencies’ decision emerging-market sovereign bonds and their impact on the dollar yield spreads. Earlier studies investigated also the behavior of Credit Default Swap (CDS) conditional to review of credit rating decision.
Another part of the literature focus on estimating the impact of monetary policy. Jonathan Kearns and Phil Manners (2005) investigated the impact of monetary policy on the exchange rate using intraday data. Jack Meaning and Feng Zhu (2011) analyzed the impact of the Federal Reserve’s Large-Scale Asset Purchase (LSAP) programme and the Bank of England’s Asset Purchase Facility (APF) on government bond yields. Eric T. Swanson (2011) present a more modern high-frequency event-study approach in order to estimate the potential benefits of FED’s QE2 by measuring the effect on long-term interest rates of Operation Twist. Don Bredin and al. (2009) focused on the private sector reaction of monetary policy, with an event study on BoE and ECB on UK and German aggregate and sectoral stock returns. For further links, see Joseph Gagnon.
Apart from these contributions, the event study methodology has found application also beyond the fields of accounting and finance. For example, in the field of development economics, Eliana La Ferrara and Massimo Guidolin investigated in 2007 AER paper the impact of the sudden end of the civil war in Angola – marked by the death of the rebel movement leader – on the stock market value of diamond mining firms.
Republishing and referencing
Bruegel considers itself a public good and takes no institutional standpoint. Anyone is free to republish and/or quote this post without prior consent. Please provide a full reference, clearly stating Bruegel and the relevant author as the source, and include a prominent hyperlink to the original post.