nd

Econometrics 2 Assignment
Submitted by
Fabio Fortis, Nikolas Aschenbrenner, Mridul Jain
Berlin 2nd February 2024

Table of Contents
3 Literature ................................................................................................................................... 2

4 Descriptive statistics ................................................................................................................. 2

5 Analysis of sorted portfolios ....................................................................................................... 3

5.1 CAPM: .......................................................................................................................................3
5.2 3-factormodel: ...........................................................................................................................3
5.3 Additional factor(s): ....................................................................................................................4

6 Fama/MacBeth regression .......................................................................................................... 5

7 Tables ........................................................................................................................................ 7
7.1 Descriptive Statistics (Sorted and Difference Portfolios) ................................................................7
7.2 Descriptive Statistics (25 Sorted Portfolios) ..................................................................................7
7.3 CAPM Results ............................................................................................................................8
7.4 Fama&French 3 Factor Results ...................................................................................................8
7.5 Fama&French 3 Factor with Momentum Results ..........................................................................8
7.6 Fama/MacBeth Results ..............................................................................................................9

8 Sources ..................................................................................................................................... 9
3 Literature
The book-to-market ratio, the ratio of a firm’s book value to its market value, is a significant
predictor of stock returns. Empirical evidence indicates that high book-to-market stocks ("value
stocks") tend to outperform low book-to-market stocks ("growth stocks") (Fama & French,
1992). This pattern has been observed across various markets and time periods, challenging
traditional asset pricing models.
Fama and French (1992) established the book-to-market ratio as a key factor in stock returns,
leading to their three-factor model. Their findings were supported by Lakonishok, Shleifer, and
Vishny (1994), who argued that the superior performance of value stocks cannot be solely
attributed to risk. Similar results have been documented in international markets (Capaul,
Rowley, & Sharpe, 1993).
The book-to-market effect contradicts the efficient market hypothesis (EMH), which asserts
that asset prices fully reflect all available information (Fama, 1970). The persistence of this
return pattern suggests that markets either misprice stocks or fail to incorporate relevant
information.
Two primary explanations exist for the anomaly: rational risk-based theories and behavioral
biases.
Risk-Based View: Fama and French (1992) argued that value stocks carry higher risk, such as
financial distress, justifying their excess returns. However, some researchers question whether
risk alone explains the magnitude of the return premium (Lakonishok et al., 1994). Behavioral
View: Lakonishok et al. (1994) and De Bondt and Thaler (1985) suggested that investor biases
drive mispricing—over-extrapolation and overreaction cause growth stocks to be overvalued
and value stocks to be undervalued, leading to their superior subsequent returns.
The book-to-market ratio remains a key factor in asset pricing, yet its persistence challenges
market efficiency. While risk-based and behavioral theories both offer explanations, no
consensus has been reached. Future research integrating these perspectives may provide a
clearer understanding of this anomaly and its implications for financial markets.

4 Descriptive statistics
The descriptive statistics table shows that higher book-to-market portfolios (Hi 10) show higher
average returns (0.044) and greater volatility (S.D. 1.662) than lower ones (Lo 10: 0.029, S.D.
1.180). Market (MKT) returns are modest (mean 0.031, S.D. 1.079), while size (SMB) and
2
value (HML) factors show minor average returns but occasional strong outperformance. The
factor of the difference portfolio (LMH) has a negative mean (-0.061), indicating
overperformance of high book-to-market stocks. Overall, value stocks outperform growth
stocks but at the cost of higher return dispersion and risk.

5 Analysis of sorted portfolios

5.1 CAPM:

Ri −Rf = α+β(Rm −Rf )+ϵ (1)

The CAPM results indicate significant differences in performance across book-to-market

portfolios. High book-to-market portfolios (value stocks) exhibit higher CAPM alphas than low
book-to-market portfolios (growth stocks). The CAPM beta values range between 0.945 and
1.277, suggesting that all portfolios have a similar level of systematic risk compared to the
market. However, the R² values vary from 0.687 for high book-to-market stocks (Hi10) to 0.953
for low book-to-market stocks (Lo10), meaning that CAPM explains a substantial portion of
return variation, but not fully.
Notably, the alpha for Hi10 is 1.207, which is highly significant, implying that these stocks
generate excess returns beyond what CAPM predicts. In contrast, growth stocks (Lo10) show
an alpha of -0.738, suggesting that their excess returns are almost entirely explained by market
exposure. This pattern is consistent with the well-documented value premium, where value
stocks tend to outperform growth stocks on a risk-adjusted basis.
The CAPM analysis confirms the book-to-market anomaly, as high book-to-market stocks earn
significantly higher risk-adjusted returns that CAPM cannot explain. The presence of positive
and statistically significant CAPM alphas suggests that market risk alone does not fully
determine returns.

5.2 3-factormodel:

Ri −Rf = α+βm(Rm −Rf )+ βs SMB+βh HML+ ϵ (2)

The alphas (unexplained excess returns) now vary between –2.673 and 1.187. Overall, the
alphas are now closer to 0, indicating that the Fama&French 3-factor model is able to better
explain stock returns and allocate them to the three factors.
3
The t-statistics for alpha across deciles vary strongly, with some of the alphas not being
statistically significant.
The value beta (HML factor) starts negative for Lo 10 (-0.333) and turns positive, increasing to
1.038 for Hi 10. This shift confirms that higher book-to-market portfolios have higher average
returns compared to low book-to-market portfolios. This also confirms the anomaly: high book-
to-market portfolios (value stocks) generate higher returns, even when controlling market risk,
size, and value factors.

The high t-statistics for value beta also validate that the increasing exposure to value stocks is
significant.
Market beta remains relatively stable across deciles, ranging from 1.068 (Lo 10) to 1.224 (Hi
10). This suggests that differences in returns are not primarily driven by systematic market risk.
The size beta increases significantly across deciles, from -0.091 (Lo 10) to 0.479 (Hi 10).
Higher size beta indicates greater exposure to small-cap stocks in higher book-to-market
portfolios. This finding suggests that part of the value premium may be explained by the size
effect, as smaller firms are often more volatile and exhibit higher returns.
The increasing value beta is central to the explanation of the anomaly. Higher book-to-market
portfolios naturally have higher exposure to the HML factor, which aligns with the essence of
the Fama-French model.
The model's explanatory power, as indicated by FF R², is consistently high for all deciles (above
85%). This suggests that while the Fama-French model explains a large portion of the variation,
some unique aspects of the portfolios remain unexplained.

5.3 Additional factor(s):

Ri −Rf = α+βm(Rm −Rf )+ βs SMB+βh HML+ βmom MOM+ ϵ (3)

The Fama-French 3-Factor Model is expanded with a Momentum factor, analyzing whether
past winners (high returns) continue to outperform past losers (low returns). Alpha values vary
across deciles, with some (e.g., Dec 2: 1.438, Dec 3: 1.085) showing positive alphas, while
others (e.g., Dec 7: -1.571, Dec 8: -0.397) are negative. Overall, alphas remain close to zero,
indicating most stock returns are explained by the model. The t-statistics also vary, with some
alphas (e.g., Lo 10, Hi 10) being statistically insignificant.

4
Market Beta remains stable around 1.0, suggesting near-market exposure. Size Beta is mostly
negative, implying a preference for large stocks, except in Deciles 8 and 9, where small stocks
outperform. Value Beta (HML) increases from Decile 1 to 9, supporting the book-to-market
anomaly.
Momentum Beta is mostly around 0, except for Hi 10 (0.852), indicating strong momentum
effects. Interestingly, Hi 10 also has a negative Value Beta, deviating from the usual Fama-
French pattern. Most Momentum Betas, except Decile 6, are statistically significant.
Overall, the model confirms the book-to-market anomaly, except for Hi 10, where a high Value
Beta was expected. R-squared values (0.876 to 0.973) show that the model explains return
variation well, outperforming the original Fama-French model and demonstrating the added
value of the Momentum factor.
The Momentum factor is added to capture the well-documented tendency of past winners to
continue outperforming past losers (Jegadeesh & Titman, 1993), which the original Fama-
French model does not explain. Its inclusion improves the model’s explanatory power, as shown
by higher R-squared values, and accounts for behavioral biases in stock pricing.

6 Fama/MacBeth regression

First stage: Ri,t −Rf,t =αi +βi Ft +ϵi,t (4)

Second stage: Ri,t =λ0,t +λ1,t βi1 +λ2,t βi2 +⋯+λk,t βik +ϵi,t (5)

With the Fama-MacBeth regression estimates, we compare the full-sample approach with a
rolling window method. The intercept, which represents the average excess return unexplained
by the factors, is higher in the full sample (1.267) compared to the rolling window (0.828), with
both being statistically significant based on their t-statistics of 4.389 and 4.901, respectively.
This result could be explained by the possibility of hindsight bias occurring in the full-sample
approach.
The market beta (betasMKT), which measures sensitivity to the market factor, is negative in
both cases, with the full sample showing a stronger negative value (-0.551) compared to the
rolling window (-0.095). However, its t-statistics suggest that market beta is not statistically
significant in the rolling window (-0.428), whereas in the full sample, it is slightly significant
(-1.716).
The SMB factor (betasSMB), which captures the size effect, is slightly higher in the rolling
window (0.183) than in the full sample (0.142), with the latter having a t-statistic of 1.447,

5
indicating moderate significance. In contrast, the rolling window's SMB coefficient has a t-
statistic of 1.899, suggesting stronger statistical relevance.
The value factor (betasHML), which reflects exposure to high book-to-market stocks, is nearly
identical across both methods, with the full sample at 0.362 and the rolling window at 0.364.
However, the statistical significance is higher for the full sample (3.377 t-stat) than for the
rolling window (3.344 t-stat), though both are clearly significant.
The final factor, betasDIFF, captures the difference portfolio as a factor. The difference portfolio
is created by going long in the portfolio with the stocks that have the largest size and highest
book-to-market ratio, and by going short in the portfolio with the stocks that have the smallest
size and lowest book-to-market ratio. The beta of the difference portfolio is higher in the full
sample (0.416) compared to the rolling window (0.311), with the t-statistics indicating that it is
strongly significant in both cases, though slightly more so in the full sample (1.570) compared
to the rolling window (1.176).
Overall, these results suggest that while the model explains a significant portion of excess
returns, the full-sample approach tends to produce stronger coefficients and greater statistical
significance compared to the rolling window approach. On the other hand, the unexplained
excess return is smaller in the rolling-beta approach, which shows that the rolling-beta approach
should be favoured.
We selected this set of test assets to capture key cross-sectional variations in stock returns,
focusing on size and value effects, which are central to asset pricing models. These assets allow
us to assess how well factors like market beta, SMB, HML, and the difference portfolio explain
excess returns. The tested specifications explain returns reasonably well, with statistically
significant factors, especially in the full-sample approach. However, some factors, like market
beta in the rolling window, show weaker significance, indicating potential instability.
The difference portfolio works well as an explanatory factor, showing strong statistical
significance in both methods. Its higher coefficient in the full sample suggests a more stable
size and value premium over time, reinforcing its relevance in asset pricing models.

6
7 Tables
7.1 Descriptive Statistics (Sorted and Difference Portfolios)

7.2 Descriptive Statistics (25 Sorted Portfolios)

7
7.3 CAPM Results

7.4 Fama&French 3 Factor Results

7.5 Fama&French 3 Factor with Momentum Results

8
7.6 Fama/MacBeth Results

8 Sources
References:
De Bondt, W. F. M., & Thaler, R. (1985). Does the stock market overreact? Journal of
Finance, 40(3), 793-805.
Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical
work. Journal of Finance, 25(2), 383-417.
Fama, E. F., & French, K. R. (1992). The cross-section of expected stock returns. Journal
of Finance, 47(2), 427-465.
Lakonishok, J., Shleifer, A., & Vishny, R. W. (1994). Contrarian investment, extrapolation,
and risk. Journal of Finance, 49(5), 1541-1578.
Capaul, C., Rowley, I., & Sharpe, W. F. (1993). International value and growth stock
returns. Financial Analysts Journal, 49(1), 27-36.

9 Packages
1. expss
2. tidyverse
3. xtable
4. readxl
5. stargazer
6. maditr
7. dyplr