Regarding random effect logit panel model

[shilpa gupta]

I have examined the micro level determinants of external commercial borrowing by non financial Indian firms for the study period 2004-2020 by applying the random effect logit panel model. Number of sample firms are 545 who have taken ECB during the study period. I have taken only those firm who have raised ECB at least twice during the study period.

Dependent variable in the study is binary as it takes value one when firm raise ECB in a particular year or otherwise zero.

There are 9 independent variables such as firm size, asset tangibility, profitability, foreign earning, foreign spending, sales growth, age, leverage, liquidity. lagged values of independent variables have been taken in the model.

I have also included time dummies and cluster at firm level.

Is this methodology correct to know the micro level determinants of ECB in India.

[Daniel Schaefer]

It's hard to say, as there usually isn't one right way to model something.

random effect logit panel model

Why not fixed effects?

Number of sample firms are 545 who have taken ECB during the study period.

I'm possibly misunderstanding, but are you saying you drop firms that didn't take (two) external commercial borrowings? Your outcome indicates the year that a firm takes the ECB, so there is some temporal variation in your variable of interest, but I'm still not sure why you would want to throw out information about firms that never take ECBs.

I have also included time dummies and cluster at firm level.

You are clustering at the firm level? I thought firms were your first order unit of analysis? If so, I think that is redundant, since the random effects model should already account for cross-sectional clustering in the standard errors by your first-order unit of analysis.

Year dummies might be okay. I guess you want to control for other non-observed differences between certain years, and you're willing to accept a certain amount of collinearity with your temporally variant independent variables?

[Joseph Coveney]

I have examined the micro level determinants of external commercial borrowing by non financial Indian firms for the study period 2004-2020 by applying the random effect logit panel model. . . . .

Is this methodology correct to know the micro level determinants of ECB in India.

If your model accurately predicts ECB in India in, say, 2026–2042, then yes; otherwise, no.

[Carlo Lazzaro]

Shillpa:

I have taken only those firm who have raised ECB at least twice during the study period.

means making up your original sample.
This choice may bias the results of your analysis.

[shilpa gupta]

sample firms include only those Indian firms who have taken ECB at least twice during the study period as i want to know the characteristics of these firms who has raised ECB

[shilpa gupta]

I have collected the firm specific data from prowess database of CMIE.
If i include all the non-financial firm which are in the prowess data base, then sample will have those firms who have raised ECB as well as those firms who have not raised ECB even once during the study period.
Will it remove the problem of sample selection bias ?
will be ok to apply random effect logit panel model to examine the micro level determinants of ECB by non-financial firm ?

[Carlo Lazzaro]

Shilpa:
I would include all the non-financial firm which are in the prowess data base and apply a random effect logit panel model (I assume that you're interested in the between-firm variation).
Otherwise, if consistent and efficient, you may want to consider the conditional fixed effect estimator available from -xtlogit,fe-.
As a sidelight, your choice should also be led by the literature in your research field, that I do not know.

[shilpa gupta]

Thank you so much

[shilpa gupta]

Hi
there are 1246 non financial firms who have raised the ECB during the study period of 2004 to 2020. In this there are 863 firms who have raised ECB at least twice and remaining firms have raised ECB only once during the study period.

I want to know the characteristics of firms who have raised ECB at least twice (at least twice, thrice or so on) during the study period. Or want to know the probability of same firms to issue the ECB again during the study period.

In this case, will it be advisable to apply fixed effect logit panel model with those firms who have raised ECB at least once during the study period. Dependent variable will be 1 when firm takes ECB in a particular year or otherwise zero. Or Should i included those firms also who have never raised ECB during the study period. want to apply the fixed effect to focus on within firms variations.

There are 9 independent variables (Characteristics) such as firm size, asset tangibility, profitability, foreign earning, foreign spending, sales growth, age, leverage, liquidity. lagged values of independent variables have been taken in the model.

Same methodology is followed by Bruno and Shin (2016) ; Acharya and Vij (2016). I am attaching these papers.

Please guide.

[Carlo Lazzaro]

Shilpa:
follow the mainstream then.

[Stephen Jenkins]

shilpa gupta The Bruno & Shin paper you attached in #9 appears to be copyright material, so I don't think you should be posting it. Much better is to provide detailed references to the original source (perhaps including DOI), i.e., no attachments. Please also read the Statalist Forum FAQ.