These are surely an amateur questions, but I am struggling with a simple cross-sectional analysis and I would like to do it correctly. Here is my case:
So I have quarterly return data for 150 pension funds for the period 2015 to 2020. Because I have too little observations for time-series, I want to start off with a simple cross-sectional analysis.
I first wonder what is the best way to basically plot the performance per fund over time in order to see the differences (excel or STATA available)?
After that I would like to figure out/explain those differences in performance among those funds. For now I have calculated like averages per quarter, over the whole period (relative performance, comparing peer group averages). I wonder if there are other easy to compute measures that I could apply with the data I have, as i.e. measures that use std dev. & betas (like Sharpe Ratio & Treynor Ratio) are not useful here (bc of little observations).
And how can I see/maybe explain differences with a simple cross-sectional regression (without taking yet into account characteristics like size etc)?
Any help is greatly appreciated!
Cross-sectional analysis (also known as relative analysis) is a comparison of a particular metric (available in any of the financial statements) of one company with the corresponding metric of another company within the same industry, or against the industry in which it operates in.
Over a 5 year period, if quarterly data on NAV's of Funds is utilized to find Quarterly returns which in Turn can be utilized to find Sharpe, Treynor, Jenson's Alpha. Not too sure if this can be called 'Scarce'
To add a bit; The quantum of data that would be available for the Risk adjusted measures would also be the data available for Regression Analysis I believe !
When conducting a cross-sectional analysis, the analyst uses comparative metrics to identify the valuation, debt-load, future outlook and/or operational efficiency of a target company. This allows the analyst to evaluate the target company's efficiency in these areas, and to make the best investment choice among a group of competitors within the industry as a whole.
See the detsiled link
https://www.investopedia.com
Fund performance can change dramatically over time. You have 5-6 years of data -- if you do 5-6 simple cross-sectional analyses and then show them in a three-dimensional graph, you will be able to evaluate the funds visually on a cross-sectional and time-series basis. You might, also, include the Industry average as a control.
Chris Wright
Dear Chris Wright thank you for your response! If you say 'doing 5-6 cross-sectional analyses' (so one per year) - how do you mean exactly? I figured while browsing, reading etc. that the term "cross-sectional analysis" is often differently used. You mean that I run regressions per each year (with performance as my Y & potential determinants as my X var.) but then I wonder how I should put that in a three-dimensional graph after, so I believe you meant sth else with simple cross-sectional analyses? Thanks a lot in advance!
Hi Marina,
What I understood about your research is that you have "quarterly return data for 150 pension funds for the period 2015 to 2020." That means you can examine only funds that were in business the entire 6 years (2015-20) or you can trim either the start year or the end year if it significantly improves your sample size -- make certain you comment on the risk of survivor bias. Each fund produces results for each of years reviewed and this is put into a three axis graph with fund, year, and performance representing each axis and then you graph the results using a 3D graphing software (e.g., Surfer, and Grapher by Golden Software are good -- https://www.goldensoftware.com/). Surfer may be overly complex and costly; The surface grid in Grapher may be sufficient and will produce incredible images for your research publications. While the result is visual rather than statistical, a New Zealand Statistician once said: If you need statistics to prove your assertions, you need more robust assertions.
I hope that this helps.
Chris Wright
- Badges
- Science topic