26. Can Margin Traders Predict Future Stock Returns in Japan?

Standardized margin transaction positions are required to be closed out within 6 months. Before the Japanese Big Bang, brokerage commissions, margin interest, shinagashi-ryo (premium charges), administration fees, and haito-chosei-gaku (ex-dividend adjustment) were determined by the stock exchange. After the Big Bang, only shinagashi-ryo and haito-chosei-gaku have been regulated. When stock loans outstanding exceed outstanding loans between the brokers and the securities finance companies, the cost of providing securities is charged to all the investors who sell the particular stock on margin; this is shinagashi-ryo. Though stockbrokers can now set their own conditions on margin transactions, they generally set the same conditions as security loans transactions. Both for standardized margin trades and for negotiation based margin trades, investors deposit 300,000 yen plus an additional 30 % or more of the contract’s value with their broker. These deposit levels are determined by cabinet office regulations and stock exchange rules. Investors are obligated to contribute additional funds to their margin account when a paper loss is incurred.

In some of the subsequent analysis we divide this sample into two sub-periods: December 17, 1994 to March 6, 1999 and March 13, 1999 to May 17, 2003. We explore these sub-periods because of a substantial increase in margin buying that started in late 1998/early 1999. This increase could indicate a change in the trading behavior of investors and may correspond to a telecommunications/internet bubble and/or financial deregulation in the late 1990s.

The TSE publishes the previous Friday’s closing number of margin shares outstanding on Tuesday. We use outstanding margin shares measured in terms of trading units that adjust for stock splits. We exclude data for weeks in the very few cases where the TSE did not make a margin announcement.

This TSE index is the TOPIX that is described below. The Nikkei 225, another well-known index, also suggests that stock prices and margin purchases move together.

Trading volume is defined as the total number of shares traded during a given week on the TSE in terms of trading units after adjustment for stock splits.

An interesting point could be made here that it is not really margin buying that is linked to market returns, but rather total trading volume. The idea is that total buying in the stock market volume displays positive feedback trading behavior. We argue that MBO uniquely captures the sentiment of individual Japanese investors, but perhaps MBO just tracks all buying on the Tokyo Stock Exchange. To test this, we also examine the relation between total buying (VOL) and the TOPIX return in Table 26.2. The table shows that there is no meaningful relation between total buying and past market returns.

The change in MBO (MSO) is calculated as the difference between MBO (MSO) this week and last week. We also use the deviation from the previous 52-week mean of MBO (MSO) instead of previous week of MBO (MSO) when we compute ΔMBO (ΔMSO). The results remain qualitatively unchanged. We only report results that use the former definition of ΔMBO (ΔMSO).

As was shown in Table 26.1, past ΔMBO is serially correlated at lag one. And as was shown in Table 26.2, past and contemporaneous returns are related to the change in margin transactions.

Since the results are qualitatively similar, we do not report results for standardized ΔMBO (ΔMSO) using trading volume.

This is true for both of the sub-periods as well. Apparently the substantial increase in aggregate margin buying from the late 1990s is not associated with a change in margin-trading behavior. We also examine the persistence of margin transactions from t = −4 to t = 4 in work not reported here. The results are essentially the same, hence we report only t = −1 and t = 1 in Table 26.4.

In sub-period analysis not reported here, the association between margin buying and future excess returns is somewhat stronger in the 1999 to 2003 period.

Order imbalance is a measure of buying/selling pressure. Chordia and Subrahamyan calculate order imbalance using an algorithm developed by Lee and Ready (1991) that classifies a trade as buyer (seller) originated if the price is closer to the ask (bid) price of the prevailing quote. We do not have bid/ask data, so we are unable to make an exact comparison of Japanese margin buying to Chordia and Subrahamyan’s order imbalance result for the NYSE.

We examined periods up to 20 weeks before and after the formation period, but no significant patterns are observed beyond 10 weeks.

One further reason for short-term sentiment is that the managements of Japanese firms are obligated to make public announcements when they anticipate that earnings in the current accounting period will be substantially different from what were originally forecast. Accounting periods were general six months in the period for our data, though Japanese firms now generally report quarterly results. These earnings revision announcements, or lack of announcements, provide important short-term information that traders might rationally use in their decision to roll over or close out their margin positions.

We also expand our estimation period up to 20 weeks before and after the formation period; however, no significant patterns are observed beyond 10 weeks.

We also conduct the same analysis for margin selling. We do not find any meaningful patterns for margin selling.

TOR is defined as average shares traded each day/number of shares outstanding. We proxy liquidity with volume in the form of TOR.

This addendum has been newly written for this book chapter.