4.1 Expected Return and the R&D Intensity
This section inspects the relation between expected return and R&D intensity.
It is sensible that firms with a higher proportion of R&D expenditure should have higher expected returns. To examine whether R&D intensity plays a role in the cross-sectional regressions of stock returns on size, book-to-market ratio, market beta, and the R&D intensity measure, jointly and separately, we apply a cross-sectional regression, following Fama and MacBeth (1973), to test for the explanatory power of these characteristics as follows:
, 1 2ln( ), 3ln( / ), 4 , 5 , 1 ,
i t i t i t i t i t i t
r = +
θ θ
ME +θ
BE ME +θ β
+θ
RD − +ε
, (1)where r is the monthly return on stock I in month t; RD is a measure of the R&D i t, intensity of RD/A or RD/BE. The regression is run monthly from January 1996 to December 2005.
Panel A of Table 6 presents the regression parameters with t-statistics of the regressions for the entire sample period. The results of returns on size, book-to-market ratio, market beta and the R&D expenditure relative to total assets separately show that each of the variables is helpful in explaining the cross-section of stock returns. The size effect is negative, but less significant, than previous studies on the U.S. and Japanese markets. The book-to-market effect is positive and is as strong as that in the U.S. and Japanese market. The market beta is positively related to average return and is very significant. The t-statistics of the coefficients indicate
that the R&D effect is not significant by itself and becomes much stronger when it is combined with other variables. All of the four variables have complementary effects in explaining cross-sectional differences in expected returns. The reported R , 2 which is the time-series average of those cross-sectional regressions, is low at the firm level for the regression with R&D alone, as well as for all the other regressions.
Panels B to D of Table 6 report the results of the cross-sectional regression of returns for three subperiods separately. Recall from Table 1 that from 1996 to 1999, the number of samples that reported positive R&D expenditure dropped slightly.
Although R&D activities might create new products and bring in more profits in the long run, the pressure from the stock market exerted great influence on the managers of these firms and pushed them to make myopic decisions. The opportunity cost of investing in R&D was obviously larger from the managers’ point of view. As a result, fewer firms were willing to invest in R&D activities and the positive R&D effect on stock returns under the normal economic environment was distorted in the bubble-forming period.
From the results of Table 5 we recall that there is a negative relationship between stock returns and R&D intensity during the bubble-forming period, which is confirmed by the coefficient of regression when R&D intensity is considered alone.
However, when we combine other variables into regressions, the coefficients turn to negative significantly. We infer that some of the firms with low or zero R&D expenditures took advantage of the rising trend of the stock market. On the contrary, the remainder of the firms with positive R&D expenditures got less returns.
After 2000, the stock market index dropped sharply due to the global recession and domestic political chaos. By September 30, 2001, the index has fallen to a low of 3,637. As most of the stocks lost value during the burst-of-bubble period, the firms that did have high R&D expenditure tended to lose less than those that had low
or zero R&D expenditure. This is reflected in the positive slope coefficients for the R&D intensity in Panel C of Table 6. For the post-bubble period, although stock prices drifted up and down without much recovery, the R&D effect remained negative and became more significant. In the post-bubble period, Taiwanese firms confronted global economic fatigue and record-breaking high prices of raw materials, meaning that it was not so easy to find a favorable opportunity for R&D investment. Those which had more R&D expenditures suffered more damage from their stock returns.
There is little special to be said for the firm size, book-to-market, and market beta across the subperiods. The size slopes remain negative in explaining expected returns, however, are not significant for all regressions for three subperiods. The book-to-market effect remains strong, and the market beta effect is significant in three subperiods. With the concern that firms of zero R&D expenditure may contaminate the result, we run the same regressions for firms with positive R&D expenditure only in Table 7. It turns out that both the magnitude and the significance of the coefficients on R&D intensity are much the same as those for the sample that includes firms with zero R&D expenditure. The R2s increase thanks to the smaller cross-sectional samples. From this we can infer that the main reason we only find a modest R&D effect on expected returns in the whole sample is actually not because firms mis-report R&D expenditure as regular investments.
4.2 R&D Intensity on the Expected Return for the Electronics Industry
As we mentioned earlier in Section 3, R&D intensity may differ across different industries. Whether a firm’s R&D intensity is high depends on which industry the firm is in. In the Taiwan stock market, the electronics industry accounts for the largest part of market weighted value. The stock performance of the electronics industry almost decides the trend of the total market index. From Table 2 and Table 3, we know that the R&D intensity for the electronics industry is highest among all
industries in the Taiwan stock market. Hence, it is reasonable to investigate the above relation for merely the electronics industry. We pick the firms classified into the electronics industry according to the classification by the Taiwan Stock Exchange Corporation and run the Fama-MacBeth regression.
Table 8 reports the results of regressions for the electronics industry only. The results again are very similar to the ones for total industries in Table 6. There is a negative relationship between R&D intensity and stock returns in the bubble-forming period and in the post-bubble period. However, R&D expenditure has a positive effect in the burst-of-bubble period. It is noteworthy in every subperiod that R2s are larger than the ones in Table 6 when R&D intensity is considered with other variables. As an explanatory factor, R&D intensity has more explanatory power for the electronics industry than for total industries.
4.3 Lag and Cumulative Effect on R&D Intensity
The lag effect and cumulative effect of R&D expenditure have recently been considered in finance and accounting fields. Given the existence of the R&D lag effect, we may argue that the current and past R&D expenditures keep releasing the benefits from the so-called know-how. Since R&D activities usually yield benefits with a time lag, the R&D effects may take time to materialize, and it is interesting to consider the cumulative R&D intensity when we examine the relationship between stock return and R&D expenditure. There have been many studies on appropriate choices of time delay in R&D effects. According to Rapoport (1971) and Wagner (1968), a range of values between 1.2 and 2.5 years is thought as an appropriate mean lag. Rather than adding lagged values of the R&D intensity, we calculate the cumulative R&D intensity as follows:
, , , 1 , 2
(CRD A/ )i t =0.4(RD A/ )i t+0.3(RD A/ )i t− +0.3(RD A/ )i t− . (2)
We now conduct a similar analysis of the regression model (1), substituting the R&D intensity with the cumulative R&D intensity. Since we need 2 years of data prior to year t to calculate the cumulative R&D intensity, the sample period is shortened. As shown in Table 9, the patterns are the same as before. Overall, both the slope and the t-statistics of the cumulative R&D intensity increase, though not necessarily for all subperiods. This confirms that R&D activities indeed have a long-term impact on stock returns.
We also do the analysis using the R&D intensity measure relative to market value to equity. The results are similar to those we got earlier. For the sake of saving space, we do not show the results again. Overall, the evidence presented in Tables 6 to 9 indicates that there is a positive relation between expected return and the R&D intensity when R&D intensity is used alone in the regression. However, when the other three variables are used together in the regressions, the R&D expenditure is negatively significantly related to expected returns.
4.4 Risks of Returns and R&D Intensity
We next turn to the relationship between risks of the returns and the R&D intensity. Such a relation is an indispensable part of our analysis. It is relatively easy to understand why R&D activities may cause the total risk of stock returns to be larger. We use the standard deviation of the return estimated from monthly data and run regressions on an annual basis. This is a practice widely adopted in the literature.
For each firm i at the end of January in each year, we calculate the sample variance of the actual stock returns over the next 12 months and denote it as σi t2,+1. Its square root is then defined as the total risk for the year from February to the next January.
The total risk is regressed on the explanatory variables known at the end of this January as the following:
σ
i t, = +γ γ
1 2ln(ME)i t, +γ
3ln(BE ME/ )i t, +γ
4RDi t, 1− +ε
i t, . (3) The coefficients are estimated in the cross-sectional regression for the years from 1996 to 2005, using the Fama-MacBeth approach. Table 10 reports the regression results of the total risk on R&D intensity. From Panel A of Table 10, we see that when used alone in the regression, the slope coefficient of R&D intensity is significantly negative. The slope becomes significantly positive after other variables are added into the regression. One possible reason is that the R&D intensity indeed has a positive effect. However, because of the positive relationship between R&D expenditure with the book-to-market ratio, the positive effect is swamped by the book-to-market ratio effect when the R&D intensity is used alone in the regression.The subperiod analysis in Panel B to D of Table 10 uncovers certain patterns for the R&D effect on the total risk of returns that are different in different subperiods.
In the bubble-forming period and burst-of-bubble period, the R&D effect on the total risk of returns is significantly positive. However, in the post-bubble period, the R&D effect is positive only when combined with other two explanatory variables.
One pattern is common across the three subperiods: the coefficient is always more positive when R&D intensity is used with other variables and less when it is used alone. Again, the reason may come from the positive relationship between R&D intensity and the book-to-market ratio.