Why do Fama French use NYSE breakpoints in the factors creation and not just aggregate all the stocks on the three exchanges and use that to create the portfolios used to create factors.
They mention in one of their papers that “microcaps”, those stocks that make up 60% of the investable universe by number only make up something like 3% of total market value. The variability in characteristics of these stocks would mean they determine most of the portfolio breakpoints when you perform sorts. The use of NYSE breakpoints prevents this.
The CRSP tape covers NYSE stocks since 1926. The Amex and Nssdaq stock data kicks in later (1960s). Therefore if you are going to use one consistent method to break stocks into market value groups it needs to be based on NYSE criteria. If you base it on all CRSP stocks there will be a break or inconsistency in the year when non-NYSE stocks are added and the number of CRSP stocks increases greatly and most of the new stocks coming in are fairly small cap. (If you plot the median (or other quantile) of all CRSP stocks MV there is a big drop at that time. Nothing has changed in the US economy, it is just that a lot of small cap and microcap stocks are joining the CRSP database that were previously not included).