Monday, November 20, 2017

CNBC, the Tax Bill, and Fundamental Fairness

CNBC, the Tax Bill, and Fundamental Fairness

I just listened to an analyst at CNBC vent about a provision of the Senate tax bill that would force taxpayers who sell stock to use first-in-first-out when calculating their capital gain tax.   The provision as it currently stands only applies to the sale of stock in companies.  Mutual funds and ETFs are exempt from this change.

The attention, focus, and emotion spent on the discussion of this provision seems disproportionate to its importance in the vast scheme of things. 

Below are my comments on this tax provision and CNBCs evaluation of this provision.

Comment One: Several other aspects of the tax bill appear to be more fundamentally unfair than this provision.   Some of the unfair tax bill provisions include – (1) lower tax rates for business owners than workers (2) taxing stipends for free tuition as ordinary income, and (3) getting rid of the interest deduction on student debt.

The financial and economic impacts of this bill on students who must borrow more to pay taxes and who can no longer deduct interest on student loans is huge, maybe even determinative of their ability to finish school.   The financial impact on FIFO for households with gains on multiple purchases of the same stock, probably less important.

Comment Two:   Currently, the tax code applies a lower marginal tax rate to capital gains than to ordinary income.   The objective of the lower capital gains tax rate is to encourage people to sell their stock so the government can obtain more tax revenue.   One could argue that taxing wages at a higher rate than capital gains is fundamentally unfair 

I can accept that Congress has chosen to lower capital gains tax rates to encourage people to sell assets so the government can obtain additional taxes.  Having given capital gains preferential tax treatment, it seems unfair to allow investors to reduce taxes by arbitrarily assigning the shares subject to tax.   FIFO seems fundamentally fair given the preferential treatment accorded capital gains taxation.

Comment Three:   Cramer’s concern is that applying FIFO exclusively to individual stock purchases and exempting mutual funds and ETFs would discourage individual stock ownership.

"The idea of owning stocks, owning them for dividend, owning them for capital gains, is a great American tradition," CNBC's "Mad Money" host said. "I guess these people don't really care about the great American tradition of owning a piece of a company."
This provision would discourage small periodic investments into stocks through a single brokerage account.  A larger investor in individual stocks could get around FIFO in the future by having two separate accounts and selling from the account with the higher or lower basis depending upon the tax strategy.

This provision if applied to mutual funds and ETFs would encourage people to split future investments across funds.  For example, an investor who makes separate investments in FUSEX (a Fidelity S&P fund) and VOO (a Vanguard S&P funds) can modify basis on capital gains by choosing to disburse from one fund over another. However, I doubt the financial advisor if employed by a firm would advocate this approach.

This diversification incentive is probably why mutual funds and ETFs became exempt.

Concluding Remark:   CNBC analysts are incredibly bright and insightful but their perspectival bias is galling.   Issues of equity and fairness are important and deserve much more attention from CNBC.   The discussion of the use of FIFO on capital gains should not fulfill the equity and fairness airtime quota.

Saturday, November 18, 2017

Median-Median Betas

Median-Median Betas

Question:   How do the betas of stocks estimated from a median-median regression compare to betas of stocks obtained from published betas, which are generally estimated with ordinary least squares regression?

Methodology:  Beta estimates for stocks are typically obtained with a least squares model where the explanatory variable is the rate of return on the market and the dependent variable is the rate of return on the stock.   The betas published on Yahoo finance were almost certainly estimated in this manner.

The median-median regression method involves sorting the dataset from least to greatest by the explanatory variable, the rate of return on the S&P 500.   The median return for both the market and firm variables are calculated for three tercile groups as defined by their ordering from least to greatest on the market return variable.

The slope of the median-median regression line is the ratio of median dependent variable top tercile minus median dependent variable bottom tercile to median independent variable top tercile to median independent variable bottom tercile.

The location of the median-median regression line is determined by the median of the middle terciles for both variables.

Here is an excellent article on median-median regression.

I previously illustrated the use of median-median regression with polling data.

I now use the median-median regression technique to estimate stock betas.

Data:   The data used to illustrate the use of median-median regression to estimate beta involves five years of monthly data from six firm.   The six firms used here are the six largest positions of six different Vanguard funds.   The fund and firm list are presented below.

Fund Description
Largest Share of ETF
Vanguard Value ETF
Vanguard Growth ETF
Vanguard Mid Cap Value
Vanguard Mid Cap Growth
Vanguard Small Cap Value
Vanguard Small Cap Growth
The same funds and firms were examined for a different purpose here.

Analysis:   The table below presents median-median regression betas and some summary statistics.

Comparison of Median-Median Beta to Yahoo Finance Beta
Type firm
Median-Median Regression Estimates of Beta
Yahoo Finance Beta
Large Cap Growth
Large Cap Value
Mid Cap growth
Mid Cao Value
Small Cap Growth
Small Cap Value

Some observations:

  • The median-median regression beta is larger than the Yahoo finance beta for all three growth stocks.

  • The median-median regression estimates are small than the Yahoo finance beta for two of the three value stocks.

  • The average median-median regression beta is 1.44 compared to 1.23 for yahoo finance.

Concluding Thoughts:

It would be interesting to conduct a larger scale study of betas estimated by different statistical methods, including least absolute error regression methods, quintile regression methods, and ARIMA models which include other macroeconomic variables. The focus of the effort might be on how betas of different types of stocks differ and on how these differences were impacted by choice of estimation method or economic model.