creating time-constant dummy in dated panel data set

[selm]

Hi,

I am currently researching the financial performance of SRI mutual funds over a ten-year period. In order to do so I am using the fama-french multi-factor model. As I am dealing with socially responsible mutual funds, they all employ a different screening strategy. As part of my research I would like to assess the impact of a certain type of screen on the constant (jensen's alpha). And this is where I run into difficulties. I would like to create a dummy variable which indicates whether a particular fund uses a certain type of screen (0 if it doesnt, 1 if it does), these are constant over the 10 year time period. The rest of my data is time-series data however, i have created a dated panel set for this. So my question is: is this possible? Can I use a time-constant dummy within a time-series equation and if so: how? I have tried to import my data from Excel but I have only succeeded in creating an undated panel set for the dummy variables and I don't know how to (if at all possible) combine this with the dated panel set I already have.

Thanks many in advance!

[EViews Gareth]

Not sure exactly what you're asking here.

Is it possible to create a time-constant dummy in a panel data set in EViews? Sure.

Is it possible to include a time-constant dummy in a panel estimation? Depends on the specification of the estimation you're doing.

[selm]

Thanks for the quick reply!

The formula I am using is this: (Rfund-Rt) = a + b(Rm-Rt) + bSMB + bHML + bMom. I'd then like to add a dummy for a specific screening category (for example Environment). The data is all time-series except for the dummy variable which is constant over time. I just dont know how to create the dummy within my data set. I have the data for the dummies structured in Excel as follows: in rows the different screening categories, in columns then the different mutual funds with either a 0 or 1 depending on whether they screen for this category 0=no, 1= yes. I just dont succeed in importing this data succesfully into my panel data set in Eviews.

Hope this clears it up!

[EViews Gareth]

Provide the files.

[selm]

I've added my eviews workfile and the excel file with the data for the dummies. Thanks!

[EViews Gareth]

How do the observations in the Excel file match with the cross-sections in the EViews file?

[selm]

They are the same funds, identified by their isin code. (if thats what you meant?)

[EViews Gareth]

Nope, not clear at all.

In the EViews file you have a panel with two dimensions. Time (months) and VAR01, which contains text labels such as _AT0000645973_P_.

In the Excel file you have two sheets. I'm going to assume you're using Sheet2, since that was what was open. Sheet two contains no time information. It contains no text that something like the labels (_AT0000645973_P_). It does, however, contain column headers that look like funds, and row headers that look like sectors. Nothing in the EViews file seems to match up with funds or sectors.

[selm]

You're right, I hadnt seen that the fund names in my eviews file were different! Ive fixed that in the Excel sheet and recreated the panel set in Eviews. (I opened the excel file and then used proc - reshape- stack in new page) Ive attached the new one. (the _ in front of every isin code gets added by Eviews itself)

Regarding the sectors you mentioned, those are the different screening categories. I was not sure how to add the time information because these variables are constant (and the other data I use is monthly). I was not sure on the structure either so sheet 1 has the categories as columns and the funds as rows and sheet2 the other way around.

[EViews Gareth]

ok, but that still doesn't completely describe what you're trying to do.

You say the dummies are time-constant.

Fine.

What is the value of the dummy for cross-section _AT0000645973_P_ for every year? How do I get that information from the Excel file?

[selm]

Ah, I understand now! Sorry, I really am a novice with structuring these kind of things (especially when Ive got no idea what to do ;))

Ive tried it as follows now, would that work? (sheet2)

[EViews Gareth]

Nope. Again, in your EViews file you have two ways of identifying an observation, time and cross-section. You say you want to create a time-constant dummy variable. That means it will take the same value for every time period, but will have different values for each cross-section. Nothing in the Excel file gives information on how to create that dummy variable. Nothing in the Excel file lines up with the cross-sections.

[selm]

Ok, i think I finally found it .. Ive succesfully added the dummy to my data set now!

However, when I add the dummy to my regression, it gives a probability of 1, so I guess that something is not entirely right still (is this the dummy trap variable?)

Ive added the excel file, it wouldnt upload the eviews one.