Testing Green Bond Announcement Effects with an Event Study

# Help interpret an event study methodology used in a famous research paper


# Help interpret an event study methodology used in a famous research paper












I wonder if someone out there familiar with reading research articles, especially finance research can help me interpret which event study methodology the author uses in her famous research article. If she uses the methodology I think she uses, its worth criticizing, as it ignores large threats that occur when dealing with multiple events. Would love to hear you opinions!

My interpretation is that she sum up abnormal returns during the event window for each event and get a pool of CARs. She then calculate the average of all CARs and the SE among CARs and divide the mean/SE and looks up the t-statistic. Traditionally multiple events studies are not conducted this way as it mitigates threats of increased variance during event window, and clustered events. I wonder if I therefore interpret it wrong and that her methods are actually better than what I can understand from reading the article. Does she use this methodology or does she apply stronger methodologies, such as ADJ BMP or BMP?

The methodology is described in the article as follows:

For full article:: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3125518

> The event study methodology examines the stock price reaction around the announcement of an event. In the following, I use this methodology to assess how the stock market responds to the announcement of the issuance of corporate green bonds. A useful feature of Bloomberg’s database is that it contains the announcement date, i.e. the day on which the company announced that it will be issuing the green bond. The announcement date (as opposed to the issuance date) is the relevant date for the event study since it captures the day when the information is provided to the market. In contrast, on the issuance date, no new information is conveyed to the market. To conduct the event study, I use the announcement date as event date (day 0). In keeping with Krueger (2015), I account for the possibility that some information may have been known to the public prior to the announcement by including the 5 previous trading days, and account for the possibility of a staggered response by including the following 10 trading days—i.e., the baseline event window is [-5, 10]. To see if there is any run-up in stock prices before and after the event window, I also consider the time intervals [–20, –11] and [–10, –6] prior to, and the time intervals [11, 20] and [21, 60] after the event window. For each firm i, I compute the abnormal returns using the market model. The coefficients αi and βi of the market model are estimated by Ordinary Least Square (OLS) based on 200 trading days prior to the first event window (i.e., the 200 trading days used in the estimation correspond to the interval [–220, –21]) using daily return data from CRSP and the daily stock file of Compustat Global. Formally, I estimate: Rit= ai + b*Rm + error where Rit is the return on the stock of company i on day t, Rmt is the daily market return, and ԑit is the residual. Market returns are country-specific. The estimated return on the stock of firm i on day t is then given by: Estimated Rit = ai + b*Rm ---> (bad mathematical formulation, but i think you get this, its standard CAPM) I then calculate the abnormal daily return (AR) of firm i on day t as follows: Actual Rit - Estimated Rit Finally, I compute the cumulative abnormal returns (CAR) for each time interval by summing up the abnormal returns within the specific time window, and report CARs for the time intervals [–20, –11], [–10, –6], [11, 20], and [21, 60] in addition to the event window [–5, 10]. The event study results are reported in Table 6 (see picture) . The sample includes all 384 issuer-day observations. For each event window, I report the average CAR as a percentage (with the corresponding standard error in parentheses). As is shown, the average CAR in the event window [–5, 10] is 0.49% and significant at the 5% level. All other inter- vals before and after this event window yield CARs that are small and insignificant, which indicates that the results are not driven by unrelated trends around the event date. The positive CARs suggest that the stock market responds positively to the issuance of green bonds.