Hacker News Reader

Comment by no_gravity | Hacker News Reader

On the first page is a nice table with the yearly performance data of Berkshire versus the S&P 500.

I wanted to see it as a graph. So I cleaned it up in VIM and then made this chart from it:

https://www.productchart.com/blog/2019-02-23-berkshire

I love how Berkshire uses honest measure to compare their performance. They could just use SP500 price index as others do and get away with it, but they choose the correct metric - SP500 with dividends included.

Berkshire is not paying dividends like most firms do and their stock value grows 2 -3 percent more than SP500 price index just for that reason.

Not paying dividends is not a way to get +3% growth. You (ideally) pay dividends when you cannot grow by using the money. Berkshire grows a lot more than +3%, and it's structural: he invests the money that other people are paying him to hold via the insurance companies. Not paying a dividend has nothing to do with the growth.

But when you pay a dividend, you lose a small % in tax, or the delay between receiving the dividend and reinvesting it. That % loss adds over time and makes dividend structures less efficient than direct reinvestment. You are correct when the business cannot grow using the money, but BRK is an capital investment company, so I don't think that's an issue for them.

Because its an investment company whose mandate is capital growth

I feel like this doesn’t lend itself well to that graph style, hard to interpret. Maybe some log scale needs to be applied to the y axis first to point out the differences.

Agree that it is not easy to interpret.

I could add a log scale chart, but since the value is the annual percentage change of market value, I think a linear scale is the right choice. But I'm not 100% sure. Would like to hear some opinions on this.

It's more interesting as log-scale cumulative returns. The fact that Buffett dramatically outperformed the S&P 500 in...uh, I think it was 1976, but it's hard to read on the chart...is mildly interesting, but the really striking figure would be "Given $1000 invested in Berkshire Hathaway in 1966 vs. $1000 invested in the S&P 500, where would you be now?"

It doesn’t have to be log scale, but you should use cumulative returns, ie add 1 to each number and calculate a cumulative product.

A line connecting the points would be very helpful.

One way to show the difference would be to sort one of the graphs - though it would mix up the years.

It doesn’t make sense to report the absolute value of the change in percentage as you are doing (in other words, values in parenthesis in the original report correspond to negative performance and you should plot them as such).

Even better, report the log((100 + percentage_change)/100) so that adding performance in different years is a meaningful operation.

Oh yeah, I overlooked the parentheses when cleaning the data with search+replace. Thanks, fixed that now.

As for the log scale .. I'm not sure if that is the right choice.

I like that this shows BRK's historical volatility relative to the index and how that's basically going away, which is indicative of the firm's age/maturity and size.

The data points are "Annual Percentage Change", but the chart isn't labeled as such.

Would you post the raw data? Just so I don’t have to clean it myself.

Sorry, I did not keep the file. I still have the 3 replacement commands that created it in my vim history though:

    %s/[^0-9]\.//g
    %s/(/-/g
    %s/)//g

Reply on news.ycombinator.com