trying to apply a chunk or two of the CFA knowledge in the real life is more challenging than I thought. In terms of determining the earnings persistence (FRA, Chapter 20, p. 319-324) I am not sure how to calculate the accruals variable in the equation:
What’s not so obvious for me is how the accruals calculated and what consists the CF-component exactly of. In other words, looking at the cash flow and income statement, which values we need to get and/or adjust?
Some suggestions:
Should we get the operating CF as given or adjust for NWC investments during the year?
Should we subtract (add) any share repurchases (issuances, bond or stock) from the CFO?
What weights of beta(1) and beta(2) coefficients should we assign? More weight to the cash component perhaps?
You must use a regression to calculate b1 and b2. To do this you must gather time-series data for both variables cash flow and accruals. You can set quarterly data but you will need at least 30 quarters because OLS estimate has a lower limit of # of observations required in the sample. About accruals. In chapter 21 you are told how to calculate accruals using either approach, the balance sheet approach or the cash flow statement approach, they are really straightforward. BS approach: accruals = NOA end - NOA beg, Where NOA (net op assets) = operating assets - operating liabilities, Where operating assets = Total assets - cash - cash equivalent - marketable securities; and operating liabilities = Total liabilities - short term debt - long term debt CF approach: accruals = NI - CFO - CFI Where NI is net income, the other two are cash flow from operations and investing activities respectively. The intention of the regression is to be able to forecast earnings using CF or accruals, if accruals explain in a high extent the earnings behavior, then the earnings have low quality. If you get a high b2 compared to b1 and b2 is very significant you may have found a company that relies too much on accruals to make its earnings grow. GL with the task!
Cool, it’s so obvious that I just wonder sometimes where the heck do I look into particularly because I used those formulas to do an initial analysis of the earnings quality… But the credit goes definitely to you since you explained it pretty understandable.
This doesn’t sound too much fun but rather cumbersome and boring… Moreover, I have never run a multiple regression but I knew I had to do this one day.
However, I embrace the challenge and thank you for the advice!
As I see the problem, you have 2 challanges, the data gathering and make the regression model to fit well. The first one is about 8-year financial statements data (I would rely only on audited statements which are at least quarterly) that you will need to tabulate to create time-series, Zzzz. Make the model to comply with OLS assumptions is hard, you will need to run many tests. After that you can conclude with confidence.
This ride will surely last at least a 40-hour work…
40-hour work? With my work-life balance this sound like at least 2-3 weeks at the very best. I feel already kind of discouraged having in mind that I haven’t run a regression so far in a statistical software package… I started yesterday with collecting the data and I struggled with the reporting practices of the company analysed. Gathering the data will be as hard as hell.
But the second part will be definitely more challenging. Reading the CFA curriculum is surely more fun… but understanding is a matter of practice.
Are you aware of any useful sources on this topic in internet which I can use?
Yeah, some experience with softwares are great on this kind of analysis, you will need help with that tho. Unfortunately I don’t know sources of practical examples, you may find some papers over there but not sure at all, sorry.