Carow Hall, Conference Room
June 06, 2014, 09:00 AM to 06:00 AM
Sentiment reveals itself in many diverse forms. Public opinion can be measured in how a populace votes on proposed referendums or the type of products an individual buys at the neighborhood grocery store. In recent years, there has been growing interest in examining and understanding the link between measures of sentiment and real economic activity. This dissertation extends both our understanding on the tendency of assorted economic agents as well as the relationship between opinion and real economic activity.
In the last few years, the online search giant Google has made aggregations of web queries available to researchers. These indexes provide broad summations of what consumers search for online, reveal relative interest across available alternatives, and provide insights into the collective inclination of economic agents. In the first dissertation paper, Exactly what does Google Know: Can What We Search for Online Predict Real Upstream Variables?, I examine the ability of aggregated web queries to predict real variables. However, unique to prior research, my approach examines real variables that are upstream from the point at which consumers make purchases. My empirical results confirm indices of online search queries can be useful in predicting the shipment of consumer electronics products from manufacturers to retailers. These results are robust across a swath of consumer electronic devices. However, my results also show that in most cases, indices of search queries do not outperform a unique and previously untested indicator which measures consumer opinion towards spending on consumer electronics products using a probabilistic survey approach. Across the six categories of consumer electronics devices examined, the median of the mean absolute error (MAE) improves by 4.4 percent when I include a search query index and 6.85 percent when I include the probabilistic survey approach. Across all device classes tested, the MAE improved with the inclusion of both the probabilistic survey index as well as a search query index suggesting both measures of sentiment add unique predictive power.
Overall the results of the first paper are consistent with the premise that the digitization and aggregation of new data streams, like that represented by online search queries, offer promising new ways to improve the predictability of real variables. The results also show well-tuned surveys using probabilistic questions can add predictive accuracy. Finally, the research opens a new vein of investigation wherein measures of consumer sentiment can successfully be applied to a subset of real variables further up the supply chain.
In the second paper, Where are the Bulls? Exploring Optimism and Uncertainty through the Business Cycle, I use household micro-level data to explore both if – and if so - how household expectations vary systematically across individual characteristics and within cohorts sharing like demographics. My specific focus is an examination of how expectations changed over the course of the business cycle from January 2007 to December 2013 – one of the most economically volatile periods in the history of the United States.
I find that individual demographic and socioeconomic characteristics influence sentiment. Namely, income, education, and age are positively correlated with future expectations while home ownership is negatively associated with expectations during the sample period. I also find that gender and race are statistically significant in explaining individual expectations while marital status and presence of children in the home do not have a statistically significant relationship with expectations .
Most importantly, I find the influences of demographic and socioeconomic characteristic influences on sentiment break down during severe economic declines, resulting in crowd behavior. An important new finding of this study was that time influenced expectations during the most recent business cycle, but only after the start of the recession. I find expectations statistically influenced by time dummies with the largest coefficients in (a) the third quarter of 2011 (following the Arab Spring), (b) the third quarter of 2010 (following the downgrading of Greek government debt to junk bond status in April 2010 and broader concerns that the financial impact would spread across Europe), and (c) the fourth quarter of 2008 (the apex of the recession and uncertainty following the collapse of Lehman Brothers on September 15, 2008).
The overarching goal was to explore and detail what, if any, demographic characteristics play a role in explaining the expectations held by individuals, how those expectations differ across cohorts of the population, and how both sentiment and the explanatory power of one’s own characteristics can change over the business cycle.
In the third paper, Today’s Keynote: do CEO presentations lead to positive abnormal stock returns?, I examine the degree to which capital markets dissect the implicit information in public addresses by CEOs of publicly-traded companies and how this in turn impacts stock prices. I looked specifically at public addresses given in the form of keynotes at the International Consumer Electronics Show (CES) – the largest annual trade event in the United States. Specifically, I examined if there are abnormal stock returns around two key dates: the initial announcement of the keynote and the keynote itself.
I find that CEO keynotes at the International CES correspond with statistically insignificant price moves in all but a few select cases. Moreover, I also find that announcements of CEO keynotes taking place at the International CES made in the months leading up to the actual keynote event likewise produce statistically insignificant price moves in almost all cases. I find this is also the case when keynotes are taken collectively as a single group.
This dissertation explores the impact and inner workings of sentiment. I ask three important questions: first, how successful – and therefore useful – is consumer sentiment at predicting movement in real variables; second, how does consumer sentiment change as the underlying environment changes, and how are individual characteristics determinant of the change? Finally, how does sentiment within the financial market community contribute to abnormal stock fluctuations?