This research thesis investigated the long-term relationship between short squeezes and stock prices. In order to do so the existing literature was presented and findings between the two variables were put forward. The initial conclusion from the literature review was that a short squeeze impacts the value of a firm over the medium and long term. This phenomenon was theorized to be caused by two drivers First, after a short squeeze, the periodic decrease in short
interest is a long-term bullish signal. Second, the higher returns during the initial short squeeze can fundamentally change the opinion of noise traders on the stock, which should also contribute to long-term positive returns.
Event study literature suggests that this increase in long-term value should be evident through a medium and/or long-term change in abnormal returns of an affected stock. This paper presented two hypotheses and aimed to find empirical evidence to confirm or refute them:
1. After a short squeeze, a stock experiences medium-term positive abnormal returns.
With the null hypothesis CAAR=0, and alternative hypothesis CAAR>0
2. After a short squeeze, a stock experiences long-term positive abnormal returns.
With the null hypothesis CAAR=0, and alternative hypothesis CAAR>0
An event study with medium- and long-term post-event windows was conducted on a selection of 11 US-stocks, which experienced a short squeeze. The analysis of the results found to reject the null hypothesis for the medium term, but not the long-term. The expected relationship between short stocks and stock prices could only partly be confirmed.
The existing literature on this specific subject is relatively scarce compared to literature on the short-term effects. It is thus valuable to investigate the subject further. The relationship between short interest and long-term stock returns is not fully understood as of yet, three confounding views have emerged in academia. The oldest one finds no relationship, while the contrarian view argues that both variables are positively correlated. Most academia seems to be of the bearish viewpoint and argues that short interest creates downward pressure on stock prices.
The descriptive statistics of this study showed that 63.63% of the sample experienced long term positive average daily returns. This simple statistical analysis should only be interpreted as weak evidence in favour of the contrarian view, as no significance testing was performed. It was not within the scope of the study to study the long-term effects of short interest on stocks, as data availability on the variable is restricted. This paper studied if there is a difference in returns before and after a short squeeze, and as such, little wisdom can be found about general price trends of the stocks through these results.
The paper argued in favour of an increase in stock returns after a short squeeze. The literature argues that a decrease in short interest should increase returns in the medium to long run (Lecce et al., 2012; Desai et al., 2002). Evidence for this relationship could only be found
in the medium term, indicating that short sellers might have covered their short positions in the short term and that subsequent decrease of short interest over multiple weeks is not strong enough to influence the returns in any substantial way. These findings are similar to little literature covering long-term returns (Vryghem, 2017)
This paper added to the existing literature a direct measurement of stock returns after a short squeeze. This analysis differentiates itself from exiting literature by using various estimation models, different composition of firms analysed and most importantly, omitting the first-day return from the medium-term and long-term analysis of abnormal returns. By doing so, this paper managed to focus on the post-event returns and added new empirical evidence to the puzzle of the relationship between short interest and stock returns past the short-term.
A robust analysis supports the conclusion of this paper. It is supported by three estimation models, incorporating an increasing amount of market risk in the estimated return variable and two significance tests: The KP-test and GRANK test. The robustness of the analysis is high, as this combination of parametric and non-parametric tests has been found to be especially appropriate to study the long-term effect on stock prices (Kolari & Pynnönen, 2011, Dutta et al., 2018). The main weakness of the analysis is not having access to daily, nor monthly short interest data and thus relying on publicly available bi-monthly data. Access to this data could on one hand help confirm the short squeezes but also help to draw a more direct line between short interest and stock returns, with the possibility to also ignore the short squeeze event element completely and focus on the short interest itself.
Furthermore, this paper was limited in its scope and analysis by multiple factors. First off, the availability of data that is more frequent than bi-monthly is proprietary. This limited the precision in the determination of short squeezes, as well as the study of short interest itself. A study with funding could acquire more complete datasets, such as the one provided by IHS MARKIT, and gain a deeper insight into the decrease in short interest after a short squeeze.
Furthermore, this study made use of the STATA command estudy to perform its event study.
Even though this allowed for less human error in the statistical analysis, it unfortunately restricted the choice of robustness checks available. Access to the intermediate regression results of the estimation models would allow for further investigations into multicollinearity and check for homoscedasticity of residuals. This could be done retroactively, but unfortunately the time constraint of writing a master’s thesis did not allow for doing so.
This paper proposes some recommendations for further research. First, further investigation into the robustness of the results would result in a more confident interpretation of the results.
Separate robustness checks on the OLS regressions of the estimation model and event study would be a good starting point. The multicollinearity assumption can be tested with a VIF test in STATA, a value of above five raises concern. The homoskedasticity of residuals can be investigated through graphical analysis. The residuals can be plotted to the dependent variable, using the rvfplot command in STATA. If the graph shows an independent and identical distribution around 0, it is reasonable to assume that the OLS conditions are fulfilled.
Another point of improvement could be incorporating Dutta et al. (2021) most recent work, one of the key improvement proposed is to use the logarithmic function of abnormal returns to further decrease scaling issues. Increasing the scope of the study by analysing a larger sample of firms would be beneficial. This could strongly contribute to more robust results and a clearer picture on how short squeezes impact returns. Finally, it seems be more effective to investigate the short interest variable directly, as have done other researchers (Frino et al., 2011; Desai et al., 2002; Deshmukh et al., 2017). In order to do so one must have sufficient research funds to acquire the necessary expensive datasets.
To conclude, this paper covered the relevant existing literature around the subject of short interest, short squeezes, and event studies in financial econometrics. It found and theorised through analysis of said literature an expected long-term positive relationship between the decreasing short interest after a short squeeze and stock returns. The paper could not fully confirm its hypothesis and only find statistically significant evidence in favour of a medium-term relationship. The study provided a reasonably robust analysis but was limited by many factors. This paper recommends future research to concentrate on data collection or data acquisition of daily short interest to investigate the short comings of this analysis further.