120 likes | 238 Views
Sentiment during recessions a discussion. by Patrick J. Kelly New Economic School At the The First International Moscow Finance Conference. Summary. Over a 101 year sample from 1905 to 2005:
E N D
Sentiment during recessionsa discussion by Patrick J. Kelly New Economic School At the The First International Moscow Finance Conference
Summary • Over a 101 year sample from 1905 to 2005: • When there is a 6 word difference between the number of negative and positive words in the typical article returns are 5.5 basis points lower across all periods. • These effects are most pronounced during recessions • 11.7 basis points in recessions, but only • 3.5 basis points in expansions
Summary 2 • Some of the higher or lower returns partially revert • Especially following positive news. • The reversion suggests the possibility this reaction to news articles is due to investor sentiment, not a rational reaction to value-relevant news content in the news. • These findings are robust to • Alternate volatility specifications (GARCH) • Outliers • Predictability of the news • Except that predictability is muted if you focus only on post open intra-day return.
Comment 1: Technical and Economic Meaning • The way the measures are developed one article can have both positive and negative words. • This ought to bias against you finding anything – BUT the pessimism measure is clear, because you classify the entire article as more (or less) negative. • Important - these effects are economically small. • 11 or 12 basis points of extra return on a portfolio with a standard deviation of over 100 basis points. • Spreads are a minimum of 6.25 to 12.5 basis points for most of this period. • Suggestion: focus on Pessimism not positive and negative
Comments: Sentiment? • The main argument that this is a sentiment story is grounded in: • That returns following news are larger during recessions than expansions • Alternative hypothesis? • If they were the same for expansions and recessions, does this mean that there is no sentiment story? • Reversion of the returns following lags of pessimistic articles • See Table 4
Evidence for Sentiment through Reversion • The evidence for sentiment is weak. • Anxiety may effect trading behavior, but a lot more evidence is needed. • Any day when the market is down – traders are going to be edgy. • If the anxiety story is correct, news should be more important following large negative returns than positive. • Suggestion: Consider interacting news and the coefficients on returns.
A sentiment story? • Not against a sentiment story – it just needs to be really well documented. For example: • Kim and Meschke (2011) look at the impact of CEO interviews on CNBC and find evidence of an effect of sentiment • Document a lack of information • The behavior of individual investors and short sellers • And eliminate dates with any kind of confounding event.
Your evidence is also consistent with rational stories • Henkel, Martin and Nardari (JFE, 2011) find the stock return predictability is strongest during recessions • They propose that the reason for this greater predictability is due to • the difficult of valuing growth options – important during expansions • The importance of cash/dividends (and financing constraints) during recessions • Your findings are eerily similar…. • Perhaps the news these columns identify is more important during recessions, • perhaps it is more tangible (about cash, not growth options)
Does the NYT identify important information? • Chan (JF, 2003) finds that extreme price moves • revert if there is no news • Persist if there is news • Perhaps this is what you are finding in your sample: • reporters do not generate news unknown to the markets, • they filter and identify news that is relevant to the markets. • Suggestion: Allow the coefficients on lagged returns to interact with your indicator for news. • This could strengthen your results and/or clarify the source of your findings.
Separating Sentiment from Rational Stories in Recession • Identify periods when sentiment is low but the market is actually in an expansion. • That way we can distinguish the more rational stories suggested by Henkel, Martin, Nardari (2011) from the sentiment based you suggest. • On the following slide consider the NBER recessions as compared to the Michigan Consumer Sentiment Survey from 1978. • Expansions do not always coincide with good investor sentiment.