Portfolio returns

[leixi li]

Hi,

I want to investigate the relationship between daily stock returns and ESG score (sustainability score) over a 10 year period. The ESG data is yearly per company and the returns are daily returns.
To do this I made portfolios with the lowest 10% in the first portfolio and the second lowest 10% in the second portfolio etc. For these portfolios, I computed the value-weighted returns based on their market capitalization. To see whether the sustainable firms have better stock returns than unsustainable firm, the variable GMB (green minus brown) is made where the return of the lowest 10% is deducted from the top 10% portfolio. The stock returns are daily and the ESG score is yearly, so every year other companies are in the top and lowest 10% portfolios. However, when I use the last row of the code provided below, the GMB variable has the same value for the whole year, so it is not sorted per portfolio. Is there a way to solve this issue and get the difference in value-weighted stock returns from the stainable firms and the unsustainable firms (GMB) per year?

/*Portfolio sorting */
gen sort_variable = ESG
sort ESG
astile portfolio = sort_variable, nq(10) by(Year)

/* compute value weighted weights */
egen weight = sum(Marketcap), by(Year portfolio)
replace weight = Marketcap / weight

/* compute portfolio returns */
gen weighted_returns = weight * Return
egen portfolio_returns = sum(weighted_returns), by(Date portfolio)

/* compute average cross-sectional sorting variable by portfolio */
egen avg_cross_sort_variable = mean(sort_variable), by(Date portfolio)

/* top-minus-bottom portfolio */
egen top = mean(cond(portfolio == 10, portfolio_returns, .)), by(Date)
egen bottom = mean(cond(portfolio == 1, portfolio_returns, .)), by(Date)
gen GMB = top - bottom

[Matthew Holt]

Does the output from

tabstat GMB, by(Year)

provide the statistic you're looking for? From what I can understand without a sample of your data to work with, it seems like this would provide "the difference in value-weighted stock returns from the stainable firms and the unsustainable firms (GMB) per year".

If not, it may help to paste the output of whatever command you're using to evaluate your results, and describe how you'd expect that output to be different if calculated correctly.

You mention that GMB has the same value for the whole year as a concern, but that seems ultimately necessary if you're trying to calculate GMB per year, as your final sentence indicates. In other words, GMB can't be sorted by portfolio if it always compares the top vs. bottom portfolio for any given year.

As a related question, are dates constant within years? Because GMB is calculated by date, and GMB keeps a constant value within year, I'd guess these variables may be redundant for these calculations (e.g., all of your dates occur on 12/31/XX). If not, perhaps you need to calculate your "bottom" and "top" variables by year, instead of by date.