Alex Conde

ECON410

Harley Davidson Analysis

Introduction

        A lot of factors go into movements in the stock market. Most people investing in a market are doing so to capitalize on some anticipated price changes in said market. However, what happens when these expectations are exceeded or not met? This is the topic of exploration in this analysis. The company in question is the motorcycle manufacturer Harley Davidson. The point of this analysis is to see how quickly news is incorporated into stock prices if expectations are not met. My initial thoughts on news incorporation into stock prices are that there would be a direct observable relationship immediately lasting for a day or two. The news that will be examined is quarterly earnings reports for Harley Davidson. This is to say that I expect the difference between expected earnings and the reported earnings to have a strong effect on the return on Harley Davidson stock for the days following the quarterly reports.

        To quantify this relationship, I first came up with the model shown in Figure I. In this model  is the drift rate, and  is the effect the earnings differences should have on the stock return. This model is good; however it does not take into account any background movement of the market as a whole. Figure II shows the same prediction equation, but no predicts the movement in the stock price relative to the entire market. It occurred to me however that this model is a bit presumptive in that it assumes that any change in the S&P 500 will directly translate to movement in Harley Davidson’s stock price. In order to remove this assumption the S&P 500 should be treated as a predictor variable itself, as the exact effect of market-wide movements on a particular stock are not known. The final equation that I will use for my analysis is shown in Figure III. In this equation  is the strength of the S&P 500’s movement on Harley Davidson’s stock price.

Data

        The first step of this analysis was to collect data. Utilizing the Bloomberg Terminals in the Learner Trading Center, I collected daily closing prices for both Harley Davidson and the S&P 500 going back to September of 2008. To obtain a rate of return for any given day I used the logarithmic formula for calculating rate of return shown in Figure IV.  I then pulled the quarterly earnings reports for Harley Davidson going back to the fall of 2008 and going up to the summer of 2016. Using Excel I coalesced the earnings reports with both S&P 500 and Harley Davidson’s return rates for not only the day of the announcement, but the day before and day following.

        After collecting all of the data and extracting the parts of interest, I transferred the data over to JMP to perform my preliminary analysis. Observing the distribution of the data revealed immediately that there were 2 distinct outliers (Fig. V). On the day of the 4th quarter earnings announcement in 2008 the S&P 500 rose 4.16%. Given that with 32 observations we can assume a normal distribution of the data, this value is an outlier at any confidence level (p = 0.0002) and will be excluded from the final analysis. Furthermore, in the 1st quarter of 2010 Harley Davidson missed its earnings projections by 96.33%. Similarly at any confidence level this observation is an outlier and will be excluded from the final analysis (see additional info at the end of the paper).

Analysis

        Fitting the remaining data to our refined model, we can see some results. Let’s first inspect the data for the day Harley Davidson’s quarterly earnings reports come out. The resulting ANOVA table (Fig. VI) shows that our predictor variables as a collective whole have a significant effect on the predicted return at the  level, indicating a strong relationship. The parameter estimates for each of the  are shown in Figure VII, and the summary of the model fit in Figure VIII. The Summary of Fit table shows us that the RSquare value for our model is 0.38, telling us that approximately 38% of the variation in stock return is explained by a combination of stock drift rate, general market movements, and earnings reports vs. earnings expectations. The Parameter Estimates table shows that the only parameter with a significant effect on stock return above the  level is the difference between reported earnings and estimated earnings. The estimate for  in our model is 0.185 telling us that for every 1% the earnings reports exceed or miss expectations we expect to see a 0.185% change in stock returns for that day. According to our analysis, neither the drift rate (if one even exists) nor the S&P 500 return for that day have a significant effect on Harley Davidson Stock prices.

        When the data for the day before and after the quarterly earnings announcements is applied to the model the outlook changes. The output tables from JMP for the preceding day (Fig. IX) and the following day (Fig. X) look similar in their Summary of Fit and ANOVA tables, but the Parameter estimates differ greatly from what was observed in before. On the days where quarterly earnings were announced, the difference between reported earnings and expectations dominated the model and was the only significant source of variation in stock returns. However, on both the day before and after the earnings announcements the S&P 500 return is the only significant source of variation, with p-values of 0.0003 for the day before and 0.0004 for the day after.

        This tells us that on the day of Harley Davidson’s earnings announcements there is a news effect dominating trading for that day, and that the return on their stock is unrelated to general market-wide movements. This effect is specific to only one day’s trading however, and by the day after the announcement general market movement is once again the driving force behind Harley Davidson’s stock return. My hypothesis was partially correct in that there is a news effect, but incorrect in thinking that the effect would last for very long. In retrospect this makes a lot of sense, since any differences between earnings expectations and reported earnings represents either an overvaluation or undervaluation in stock prices, and once this is shown the stock price will be corrected by trading very quickly.         

Images and Equations:

Fig. I

Fig. II

Fig. III

Fig. IV

Fig. V

Fig. VI

Fig. VII

Fig. VIII

Fig. IX

Fig. X

Fig. XI (this is the full JMP table I used for all of my analysis)

Additional Thoughts:

        Something that occurred to me while inspecting the data collected was the seasonal price variances. I think it would be interesting to see if the severity of observed weather conditions vs. expected weather patterns exhibits the same sort of behavior that is seen with the quarterly earnings announcements.

        On the topic of the 1st quarter earnings announcement of 2010 where Harley Davidson missed expectations by 96.33%, I think there could be an explanation. Recall that the winter of 2009-2010 was the year that most of the United States experienced record breaking snowfall. It follows logically that if large areas of the country see much higher than normal snowfall that sales of motorcycles might drop. I imagine that lower than average earnings were expected, but no one knew the severity of how low they would actually be.