- Alternative Data
- Market Data Approach
Nikkei demonstrates how POS data may be applied to equity investment
Factor investing has been the subject of research and experimentation, especially in equity markets, and has been widely used in the asset management world. Especially in recent years, with the improvement of computer performance and the development of machine learning techniques, a wide variety of data that were not much used in the past have been applied in factor construction. Such data is called "alternative data"—as opposed to "traditional data"—and includes consumption data such as POS (Point of Sales) data and credit card transaction data; text data such as news; foot traffic data such as smartphone location information; and images such as aerial photographs. Here we will focus on POS data.
POS (Point of Sales) refers to information collected from POS systems, where transactions between a business and its customers occur. This data typically includes details such as products sold and their producers, categories, prices, quantities, and transaction dates. In this study we used Nikkei's own POS data, which is mainly collected from supermarkets, to construct a variety of factors. As POS data is often used for sales analysis and inventory management, we first forecasted quarterly sales growth rate (%). As quarterly sales are often affected by seasonality, the model used was a time-series model with an unobserved component (seasonality). The sample data used was from the first quarter of 2008 to the fourth quarter of 2023, with the period from 2008 to 2018 as in sample, and 2019 to 2023 as out of sample. Extended rolling was applied.
Figure 1. Comparison of Kao's actual quarterly sales growth to our forecast
Figure 1. shows the results in the case of Kao Corporation (4452) as an example. The blue line indicates actual sales growth rate (%) and the orange line shows our forecast. The orange line tracks the blue one closely, and the correlation between them is 0.94 in the out of sample. The results we obtained for other firms are similar to this one, which demonstrates that Nikkei's POS data is very useful for forecasting companies' sales.
Next, we applied our forecasts to actual equity investments. As Nikkei's POS data has a time lag of just a couple of days, we are able to forecast a firm's quarterly sales a few days after the end of a quarter, which is roughly a month before the actual earnings release date. We constructed an earnings surprise scenario well ahead of the actual earnings release date by comparing our forecasts with the analysts' consensus forecast (based on sales), then conducted a back test. The universe was limited to 81 stocks which the Nikkei POS data above matches well (mainly foods at 57% and raw materials and chemicals at 28%, as per TSE 17 industry classifications). The back test period was from January 2019 to March 2024. The earnings surprise was defined as (POS sales forecast - QUICK consensus)/QUICK consensus and the surprise was calculated on a quarterly basis. We used the pooled sample for the back test. The top 5% of surprises in earnings terms were defined as Top, and the bottom 5% as Bottom, with the holding period set from six business days ahead of the earnings release date to one business day after the release. Idiosyncratic returns, not total returns, were used for the performance measure to exclude market and style effects.
Table1. Performance of earnings surprise strategy
Table 1. shows the performance of each portfolio (average), where Long/Short is the top 5% minus the bottom 5%. The first row shows the results of the earnings surprise strategy, and the second row shows cases where our forecasts were 100% correct. The comparison demonstrates that although the former posts 4.2%, a little lower than the latter at 5.1% in terms of Long/Short, the earnings surprise strategy based on Nikkei's POS data performed well. Kao Corporation is a manufacturer of household and chemical products. The Company produces cosmetics, laundry and cleaning products, hygiene, fatty chemicals, edible oils, and specialty chemicals.
Thus far we have focused on sales in the POS data. However, Nikkei's POS data has a range of other variables. We constructed about twenty-five unique POS factors related mainly to a firm's fundamentals and they are now part of our SMACOM* product. Among them are the product price change factor, the product price stickiness factor, the product and category concentration factor (based on the Herfindahl-Hirschman Index), the volatility of the number of products (a proxy for scrap and build cost), the market share factor, and factors related to new and disappearing products. The back test results with these factors demonstrate a strong cross-sectional return prediction power. Furthermore, these factors have a low correlation with traditional quant factors.
It is expected that the use of alternative data will accelerate in the asset management industry. This article indicates that Nikkei's POS data can be a unique source of new factors that are related to a firm's fundamentals and those factors can be effective in terms of alpha and diversification.
*SMACOM: https://www.ftri.co.jp/product/smacom/en
SMACOM free trial (below):
https://www.ftri.co.jp/eng/index.html#company
Disclaimer (PDF file):
https://www.ftri.co.jp/eng/pdf/SMACOM_Disclaimer-EN.pdf
Nikkei FTRI
Nikkei FTRI is a member of the Nikkei Group that works with data analysis technology. We are recognized for the high quality of our analytical and modeling techniques, which utilize both traditional and alternative varieties of data.
See More