Trending Content
| +81 | Q&A - Buyside Equity Research Analyst/PM | 29 | 6d | |
| +32 | How the heck do you learn all of this stuff? | 10 | 1d | |
| +28 | My experience on the sell side | 20 | 2w | |
| +24 | Clues in Financial Reporting Analysis | 7 | 2d | |
| +17 | Am i fairly compensated? | 9 | 3d | |
| +14 | Data Science to PM | 8 | 1w | |
| +14 | Heard they don’t model at Oppenheimer | 10 | 2w | |
| +12 | ER - Remote Jobs? | 8 | 6d | |
| Equity Research at Bank vs IG, High Yield & Leveraged Loans Research at Asset Manager | 4 | 1d | ||
| +11 | Who uses sell-side Strategists? Are they valuable? | 4 | 1w |
Career Resources
Wait what? You can't be serious. You're going to regress revenues to figure out if there's a correlation to future revenue?
i.e. use past data to figure out future revenues?
A professor suggested I use a regression to forecast revenues; try to find historic variables that revenues were highly correlated to. Variables that ultimately arrive to a statistically efficient equation with high R values, sufficient residuals, and low p-values.. and at the end of the day make intuitive sense. However, using this equation to forecast means you have to forecast each variable which to me adds a lot of risk and uncertainty because the forecasted value depends on your assumptions. This takes a lot of time... and it seems that the incremental benefit of using such a method is minimal at best. Thoughts?
Well your professor is a fucking idiot
Go fuck yourself. This was actually a fairly legitimate question, especially in regards to theory vs. practice. CFA curriculum actually includes regression as a forecasting measure.
IBD does not forecast using any sort of regression methodology. It's alot more qualitative.
If I was thinking from your professor's vantage point, even if you were trying to project sales based on some historical variable (say "X"), it would still require a forward-looking assumption on X for every projected year; so essentially, your correlation will be a constant mutlplier of function X.
Nobody does this. They go much more in depth. For example, you could say that in the past when GPA growth has been X1% and the unemployment rate has been X2%, sales have grown at Y% and create a regression that way.
But the actual way people would approach this is by breaking it down further than that. If GPA is expected to increase by X1% next year, where is that growth coming from? Is it increases in consumption spending, government spending, fx transactions? And within that, what specifically? Will it even impact our industry, or our company?
If unemployment is supposed to decrease from 8% to 7% over the next year, what is the cause of that? Are people leaving the labor force? If people are getting hired, who is it? Skilled, unskilled?
The end result of all of this is they want to figure out how the target market for their company's product might change. You need to have actual numbers and estimates, i.e. I think the potential market for good XYZ will increase by 200mm people, and at the current market penetration rate the company should capture 25mm of these, which will increase revenue by $500mm
The analyst that I used to work for told me that in industry (he used to work at Unilever) they used regression to forecast things like future air travel passengers and the company would base its budget/targets off of that. He initially wanted to use it in the ER reports but was shut down by product control, as they don't feel that it's legitimate.
The effectiveness of using a regression model will depend on the sector. For an O&G company its obvious that the top line will depend on the price of oil. The problem with regressions as I see it is that you need at least 40/50 data points for a normal distribution bell-curve, so maybe 40/50 quarters, which in itself is not hard to come by for most S&P 500 companies. The main issue is nonstationary data or 'breaks' in the data where different parts of the time series have different statistical properties. This usually happens due to changes in the sector: technology or regulatory changes being the major factors. I would aboid using regressions unless you know how to test for breaks and make adjustments.
This is pretty key. Unless you are building out an internal operating model and have monthly or weekly financials, it will be very difficult to come up with enough data points for an effective regression, especially if you are observing seasonality. With quarterly financials it's too difficult to get a significant N without hitting breaks.
GM,
Did you even go to college bro? Do you even work bro? Regression analysis can be used to forecast revenues and pretty much anything else for that matter (at least in theory).
Gulf, anyone who actually uses regression analysis uses complex software and/or detailed ANOVA tables. In practice, everyone relies on qualitative bullshit and other people's opinions.
WSO really making the demographics of this board look good
http://www.econjobrumors.com/topic/reg-monkeys-on-wall-street
Yes, we use regressions to back test our model v.macro assumptions - i.e. does our organic growth fcast make sense given the GDP fcast. I know that the buy side uses a similar approach looking at companies in our sector - macro variable multipliers
Placeat magnam ipsam qui fugiat occaecati velit quas. Voluptatem dicta magni saepe tempore sit facere natus. Id ut illum quas sunt nihil eius dolor tenetur. Voluptatem quae et incidunt voluptas. Adipisci corrupti libero praesentium molestiae deleniti et ut est.
Voluptatum veniam ea rem incidunt. Tempora quis aut commodi at sunt ratione ullam. Suscipit omnis provident modi at velit dolor et illum. Sed sint aut numquam aperiam. Consectetur odit vero voluptatem nisi porro occaecati.
Possimus eius distinctio quo in aliquid est. Quia molestiae consequuntur ducimus voluptatem molestiae nihil velit.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...