One more thing. In Yardeni’s pdf, the squigly lines over the CNY series are not our CurrY & NextY. They are estimates for the specific year. That is also the correct way to do it. I don’t think they are worth the trouble to re-create since the CNY is enough IMHO
No the CNY has not changed. You can see the new one below or in this google drive spreadsheet:
As you can see the CNY is in between the CY & NY , as it should be , and the new one seems a bit smoother. Probably not enough to justify changing it. Using quarter # to slot in the CY is a perfectly valid way to do it (Compustat does it with Q#).
Estimate data is very spotty before 2000. Reuters was the same.
The SPEPSCY jump in 2010 is legit. It’s simply CY switching from 2009 to 2010 . This is precisely why CNY should be used instead.
Make sure you exclude berkshire with: UnivExclude(“BRK.B”)
We have not been able to resolve the issues with berkshire switching “primary” stock from A to B on the Compustat side while CapIq maintained A as the “primary” issue.
Is it possible to show us a sample #SPEPSCNY in “Custom Series” so we can adjust it if necessary? For example, I want to see some sector-specific EPSCNY.
But please explain how P123 calculated the last value (2/6/2016) of EPS CurrY = 118.13975 in the FedModel.
That would be the with the Friday 2/5/2016 data.
I one uses the custom series MySPEPSCY to calculate this value then it becomes 113.84. There is a huge difference here.
For the previous 4 weeks the difference between the FedModel data and the custom series values was much less.
#SPEPSCNY: Calculated using a blend of the Current and Next Year estimates for the S&P500 stocks. The weight given to the CurrY and Next Y depends on which quarter was the most recent. For each SP500 stocks the following is calculated:
BlendCNY = if Q4 then (CurFYEPSMean)
if Q1 then 75% (CurFYEPSMean) + 25% (NextFYEPSMean)
if Q2 then 50% (CurFYEPSMean) + 50% (NextFYEPSMean)
if Q3 then 25% (CurFYEPSMean) + 75% (NextFYEPSMean)
The SP500 Blended EPS is then computed as
num = Sum { BlendCNY (i) * Shares (i) }
den = Sum { MktCap (i) }
Geov, 113.84 is coming from you clicking “Totals” right ? There’s a issue with clicking “Totals” on a Monday. It’s uses the benchmark last date available to set the “as-of-date” for the estimates, but that’s a Friday, so it’s missing this weekend data. The backtest, or time series generation, is correct though.
Long story short… we’ll investigate “Totals” so that it’s not confusing. It was always meant to be just a way to spot check data.
Marco, Thanks a lot for this explanation, very much appreciated. I noticed that the graph showed the higher value at the end, in line with the FedModel data.
Not sure how P123’s SPEPSCNY is computed or if it is computed correctly.
Since P123’s EPSCurrY and EPSNextY are both going up right now, why do we have falling SPEPSCNY? See attached chart.
CNY represent “the year ahead”. It is falling now because more & more companies are reporting Q4, which results in 1/4 of NextY being dropped off, and more & more CNY being the same as CurrY. Since NextY is usually bigger than CurrY , during the first few month of the year when most companies fiscal year ends, CNY moves towards Curry resulting in a decline. A real decline, not artificial.
Marco, I know you have gotten beaten up about this. But I wanted to say thank you for being open on your inputs and rational on this topic. It made me rethink a timing strategy I have that was using CY and I have redone it completely using CNY. My overall strategy got better because of this re-think.
Since the Jan and Feb values of SPEPCY are unreliable, one should not use those values in a model, especially when there is a down-market present. I got good results with SPEPCY in timer models when restricting the model from going long during the period Jan-Feb while down-markets exist, indicated by sma(50,0,#bench) < sma(220,0,#bench).