How to construct 10 portfolios from 200 stocks

[Richard Mawu]

I have 200 stock returns. I want to construct 10 portfolios, using delices sorted by the stock's volatility. Help on how to write a loop for this will highly be appreciated.

Thank you

[Clyde Schechter]

-help xtile-

[Attaullah Shah]

You can try the following

egen portfolios = xtile(volatility), nq(10)

If you want to do that each year, then

bys year: egen portfolios = xtile(volatility), nq(10)

[Nick Cox]

I don't know what delices are. Note that Attaullah's solution requires download from SSC, i.e.

ssc inst egenmore

[Richard Mawu]

Thanks for the quick response, I already checked xtile function and I have read comments on similar post in the forum. However, I can’t tell I still feel there is a mistake with the code I came up with.

***Firm is Firm1 to Firm200 daily stock return
***V is volatility of each stock hence V1 to V200
***I want to create 10 portfolios sorted on the volatility variable
*** I use this code but the result I get is strange and am not sure if the portfolio is sorted by Firm1 on V1, Firm2 on V2 and so on

************************************************** ************************************
sort Firm* V*
foreach X of varlist Firm* {
egen portfolio`X' = xtile(`X'), by (V*) nq(10)
}

************************************************** ***********************************
Thank you

[Clyde Schechter]

No, you can't do this in wide layout. The earlier responses assumed you had 200 observations each containing a firm and its volatility: that's the long layout. You have wide layout: 200 Firm variables and 200 volatility variables. So you have to go long:

gen long obs_no = _n
reshape long Firm V, i(obs_no) j(_j)
xtile decile = V, nq(10)

More generally, most things in Stata are easier (sometimes even only possible) in long layout than wide. Having data like this arrayed as a separate variable for each firm and each volatility is a recipe for a difficult mess.

[Richard Mawu]

The length of the data is 1000 daily data. When I try your code I get an error

gen long obs_no = _n
reshape long Firm1 V1, i(obs_no) j(_j)
(note: j = 0)

variable Firm1 already defined
r(110);

xtile decile = V1, nq(10)


The data has Firm 1 to Firm 200 (returns) and then V1 to V200 (volatility of each firm)

Date	Firm 1	Firm 2	Firm 3	Firm 4	Firm 5	Firm 6
4/01/2010	0.047164	-0.09991	0.007698	-0.03226	-0.03226	-0.00306
5/01/2010	0.047063	-0.1054	-0.00847	-0.0435	-0.0435	-0.01777
6/01/2010	0.042946	0.024919	0.006855	0.027033	0.027033	-0.00602
7/01/2010	0.04271	-0.05712	0.023591	0.01817	0.01817	-0.03251
8/01/2010	0.040565	-0.04292	-0.04417	-0.02588	-0.02588	-0.03642
9/01/2010	0.037597	-0.00684	0.022761	-0.00361	-0.00361	-0.00493
10/01/2010	0.03731	0.041008	-0.00167	0.013533	0.013533	0.00314
11/01/2010	0.036982	0.037516	-0.02636	-0.01258	-0.01258	-0.03
12/01/2010	0.036849	-0.03332	-0.01688	-0.02125	-0.02125	-0.01708
13/01/2010	0.036478	-0.04201	0.000312	-0.02392	-0.02392	-0.01279
14/01/2010	0.035534	-0.02668	0.000742	-0.00341	-0.00341	0.007816
15/01/2010	0.033753	0.0432	-0.00299	-0.00527	-0.00527	0.005277
16/01/2010	0.032578	0.023811	0.031447	0.013726	0.013726	0.009029
17/01/2010	0.032346	-0.01103	-0.00341	0.001877	0.001877	-0.02545
18/01/2010	0.031923	-0.04964	0.008454	0.003861	0.003861	-0.00261

Snapshot of part of the data.
So basically what I want to do is sort Firm1 on V1, sort Firm2 and V2 to Firm200 on V200. Then create 10 portfolios from that.

[Clyde Schechter]

Well, for starters you tried

reshape long Firm1 V1, i(obs_no) j(_j)

but my advice was

reshape long Firm V, i(obs_no) j(_j)

The absence of the 1's is important for this command to work properly. When working with code, you have to follow the details exactly.

That said, now that I see something of your data, I no longer understand what you want to do. I don't understand what you mean by sorting Firm1 on V1, etc. Nor, given this layout of the data with daily information, do I understand what you mean by a portfolio based on deciles of volatility.

By the way, in the future, it is best to post example data using the -dataex- command, which you can get by running -ssc install dataex-. It is easy to use, and the instructions are in its help file, -help dataex-. If I had understood what you are trying to do, I might have wanted to experiment with your data. But given the way it is posted, I really can't. For example, there is no way I can tell from this whether your Date variable is stored as a string or is a Stata internal format date (or, ugh!, some simple integer variable with value labels). Without knowing that, I could end up showing you code that would work in one of those situations but fail in your actual situation, thus wasting both of our times. So please use -dataex- from now on. It guarantees that anybody who wants to help you can construct a completely faithful replicate of your data with a simple copy/paste operation. (By the way, if you do post a data example in your response, please be sure to include some of the V variables as well.)

[Nick Cox]

delices is a typo for deciles. I just didn't see that. I wasn't being ironic (on that occasion).