README_Quant.md provides quantitative context behind the code. If you want to learn more about the technical elements and usage patterns, visit README.md.
At my old quant shop, every financial data-item you could reasonably want was sitting neatly for you in a SQL table. In a few joins, you usually had what you wanted.
Fast forward to now: I got nothin. Most research projects require a few fundamental data-items just to get started: date-stamped identifiers, asset returns (prices/dividends), industry constituents, and broad market indices. Traditional Index vendors like MSCI and Russell charge exhorbitant amounts of money for a history of index constituents and other basic data items.
The core idea behind this project to use ETFs as proxies for broad market/sector indices.
By design, each ETF tracks its index very closely, and they're also freely available through most all company websites (iShares, SPDRs, Invesco, Vangaurd, etc.). In this project, I extract data from iShares because of its comprehensiveness and ease-of-access.
The list of all 300+ available iShares ETFs can be found here.
Here are the top 10 by AUM:
A sample file looks like this.
- "Universe" histories - to the extent that your model is cross-sectional (comparing stocks across a distribution, at a given point in time) it's important to know what companies existed when.
- Benchmarks - comparing (and optimizing) your strategy against a low-cost ETF is as good a benchmark as any
- Stock-Level Exposures to Sectors - meta-data within each holding file contains sector assignments, which are critical inputs to model construciton and performance attribution (industries explain roughly %X percent of cross-sectional variance)
- Understand Market movements - with over 300 ETFs, iShares has nearly every corner of the market covered; tracking returns of different representative categories helps describe micro market movements.
- Id-mapping: point-in-time SEDOLs, CUSIPs, ISINs, Tickers, etc.