Event Study with Latent Variables

Frequently asked questions about PLS path modeling.
Post Reply
Christian1988
PLS Junior User
Posts: 6
Joined: Mon Dec 12, 2016 12:59 pm
Real name and title: Christian

Event Study with Latent Variables

Post by Christian1988 » Thu Mar 15, 2018 5:07 pm

Dear All,

I am a doctoral student and focus my work on corporate M&A. I would highly appreciate any input on my question!

I try to measure the effect of a Type II multi-dimensional second-order construct (reflective-formative type) as an independent variable on cumulative abnormal returns (CAR) as the dependent variable (DV). More precisely, my DV is the stock price reaction to M&A announcements.

Hence I try to combine survey data (latent variables) with stock market data. My controls largely consist of accounting data.

(1) Is my understanding correct, that I should calculate CAR for each transaction in STATA as the calculation in STATA more convenient and then I integrate the CARs in my dataset and calculate the relation between the IV (latent variable) and DV (CARs) in SmartPLS?
I also would integrate control variables (leverage, Tobin'sQ, Market Cap, etc.) in SmartPLS and would treat them as single item measurs (same for CARs). Would this be the right way to go along?

(2) It it correct that another possible way is to estimate the latent construct scores in SmartPLS and incorporate those in STATA to calculate the multivariate regression for my event study there (latent construct scores as IV and CAR as DV)? Is this way possible / as clean as the other one as well?

Thank you so much for your help!

Kind regards,
Christian

jmbecker
SmartPLS Developer
Posts: 879
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Event Study with Latent Variables

Post by jmbecker » Sat Mar 17, 2018 6:03 pm

Yes, I both ways should be possible and should ideally give you very similar results. Yet, they are not equivalent so don’t expect exactly the same results.
I would probably prefer version 1 if you would only estimate a normal regression in STATA. If you want to go with fixed or random effects then I would prefer version 2.
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Ja ... v=hdr_xprf
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de

Christian1988
PLS Junior User
Posts: 6
Joined: Mon Dec 12, 2016 12:59 pm
Real name and title: Christian

Re: Event Study with Latent Variables

Post by Christian1988 » Fri May 25, 2018 7:15 pm

Dear Dr. Becker,

thank you very much for your answer.
I use option II (calculating latent construct scores / measurement model) in SmartPLS and calculating path coefficients / structural model in STATA) as I have a fixed effects regression.

Are there any published papers that follow a similar approach (using both SmartPLS and STATA for calculations).

Thank you so much for your help!

Kind regards,
Christian

Post Reply