Evaluating Predictability of Share Buybacks to Inform Investment Decisions: A Case Study

It is generally accepted that share buybacks tend to boost stock prices, so the ability to predict share buybacks should be highly useful in investment decisions. We evaluated whether we could predict corporate buyback announcements, using the Share Buybacks Data provided by QUICK, a financial information provider in Japan.

Demonstrating that Share Buybacks Boost Stock Prices

We examined changes in stock price before and after share buybacks using data from all publicly traded companies in Japan that repurchased their own shares between 2012 and 2023. The below graph shows the average percentage change in stock price of companies that carried out buybacks each year. It illustrates the changes in stock price over the three years before, and after, the year of their buybacks (T=0). The bold red line shows the overall average of the stock price changes.

For example, the "Share Buybacks in 2012" line shows the percentage change in the stock price of each company that announced a buyback in 2012 over the three years before and after the announcement, and averages these results across all the companies' stocks.

The overall average (the bold red line), shows that while stock prices show an overall upward trend, the rate of increase is steeper following year of the share buyback. This indicates that buybacks tend to increase a company's stock price.

Evaluating the Predictability of Share Buybacks

If share buybacks help increase stock prices, then the performance of an investment portfolio could be improved by including companies that are expected to repurchase their own shares.

Therefore, we attempted to predict buybacks using the Share Buybacks Data provided by QUICK.

Methodology

Machine learning was applied at the end of each year to predict the likelihood of a buyback occurring in the following year.

The objective variable was the probability of a share buyback occurring in the following year. The explanatory variables used were Share Buybacks Data (e.g., past buybacks); financial data (e.g., sales, equity ratio); stock prices and indices (e.g., log market capitalization); and other data provided by QUICK, aggregated as of year-end.

For example, to predict the likelihood of a share buyback being conducted in 2024, the following steps were taken: (1) Aggregate the learning data (explanatory variables) as of the end of 2023; and (2) Use machine learning to predict whether a buyback (the objective variable) will take place in 2024.

Using this forecasting method, we examined the accuracy of buyback predictions each year from 2013 to 2022. We calculated the Area Under the Curve (AUC), which is used as an indicator to evaluate the model and verify its accuracy. The closer the AUC value is to 1, the more accurate the model is.

The resulting AUC values were high, averaging about 0.8 for each year. This demonstrates that share buybacks in the following year can be predicted with a high degree of accuracy in all years, and that past buyback and other financial data can be used to predict future buybacks.

Investment Strategy Using Share Buybacks Data

Using the QUICK data, we found that the likelihood of share buybacks can be predicted with a high degree of accuracy, we went on to devise an investment strategy using this prediction and evaluated its performance.

In this study we tested the following investment strategy:

1. At the end of each year, calculate the probability that each stock included in the study will announce a share buyback in the following year.

2. Rank the relevant stocks according to the probability of a buyback, and sort them into top, middle, and bottom groups based on the probability of a buyback.

3. At the end of the year, purchase stocks in the group with the highest probability of a buyback, with the assumption that the stock prices of those companies will rise.

4. Sell the relevant stocks at the end of the following year.

The below graph shows how this investment strategy performed from the end of 2012 to the end of 2022 (the cumulative excess return against TOPIX over time). The stocks included were those listed on the Prime Market of the Tokyo Stock Exchange (excluding non-Japanese companies and financial companies).

The results show that when the portfolio was composed of stocks with a high probability of share buybacks, the cumulative excess return ended up at about 30%, and the annualized average return remained steady at around 3%.

Use of Share Buybacks Data in Investment Decisions

In this study, we evaluated the predictability of future share buybacks based on a company's past buyback data and financial condition. We found that buybacks can be predicted with a high degree of accuracy and these predictions can be applied to investment strategies. The Share Buybacks Data is extremely useful in making investment decisions.

This study used a simple method of aggregating stocks with a high probability of share buybacks and examining their performance. There may be room for further improvement, such as in the selection of learning data and the choice of stocks for inclusion. For example, certain buybacks may have less or more impact on stock prices. It may be possible to improve investment performance by identifying and excluding stocks whose prices are less influenced by share buybacks.

The Share Buybacks Data introduced in this case study are available from QUICK via API. In addition to information on whether or not companies conducted share buybacks, the dataset contains a wide range of other information, such as each company's progress in share buybacks. We highly recommend it for use in the selection of stocks.

Share Buybacks Data on QUICK Data Factory
https://corporate.quick.co.jp/data-factory/en/product/data029/

Back