Adjusting Asset Returns for Dividends in Replication Regressions

Summary

The document asks how to include dividends when regressing the returns of two assets against a target asset, and how to interpret the regression for return and volatility estimates. Its concrete guidance is to avoid an ex-dividend price jump by approximating dividends as continuously paid and adjusting asset returns accordingly. Dividend amounts can be estimated from historical data or modeled as proportional to asset prices.

The response offers a simplifying treatment rather than a full regression procedure. It does not explain how to calculate annualized return or volatility from the regression, how to construct a replication portfolio, or how coefficient standard errors differ from model-level uncertainty. The adjustment may be useful when a continuous dividend approximation fits the assets, but the document gives no empirical comparison or detailed assumptions for applying it.

Key ideas

• A continuous dividend approximation can reduce ex-dividend date jumps in return series.
• Dividend adjustments can be estimated from historical observations or modeled in proportion to asset prices.
• The response does not specify how to annualize regression-based return or volatility estimates.
• It leaves the distinction between coefficient standard errors and overall model uncertainty unanswered.

Tags

# How to run an asset replication regression?


# How to run an asset replication regression?












I am doing extensive research on portfolio replication and was hoping to get some help with some problems I am encountering.

I am running a regression between 2 assets that I believe replicate another asset well. For example, let A be the asset we are replicating and let B and C be the assets we will use to mimic A. I want to run a regression on the returns.

How would I run a regression such that I can include the returns on dividends paid on each of these assets. Would I just add in the amount of the dividend paid to the stock price on the day the dividend is paid? Would I then calculate a return based on those numbers?

Also, how would I use the regression output to find annualized volatility and return on asset A? What about assets B and C?

Finally, what would be the difference between the standard error of each of my variables (B and C) vs. the standard error of the model as a whole?

## Answer by Alexey Kalmykov (score 3)

https://quant.stackexchange.com/a/2853

Regarding the dividends:

In order to avoid jumps on ex-dividend date, you can make the simplifying assumption that dividends are paid continuously and adjust the returns of the assets. The size of dividends could be estimated from historical data or can be set proportionally to the asset price.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.