Stata Tutorial: Instrumental Variables with Panel Data-Basics

Basic commands and tips on using the 'xtivreg' command in Stata with fixed effects in panel data. We use psid example data to discuss endogeneity bias in a composite error model.

[@elodiegradlife6904]

2:52 tsset id t to set panel data 7:33 xtivreg Y1 (Y2=Z) X, fe 9:39 reg Y2 Z X (first stage), look for near 10 F and significant relation between Z (exogenous var) and Y2 (endogenous var) 10:20 xtivreg lwage (wks=ms) exp, fe (second stage), then to look for differences of coefficient of Y2 10:54 xtreg lwage wks exp, fe , find out differences so to affirm endogenous existed

[@flavianepomuceno7465]

I am a big fan of your videos, Mike! Thank you for this super didactic explanation.

[@samchipwanyanemapfupa4061]

Thank Mike! you are a very good teacher :) Zim approves

[@mikejonaseconometrics1886]

You are welcome!

[@alexanderopoku6835]

Thank you, Mike. but I am running IV fixed effect and am having a similar issue you had. we the fixed effect coefficients were positive and significant but the ivfixed effect coefficient is negative and insignificant. how can I solve this problem?

[@ChangYu-j7o]

Hello, the video is very helpful. But I have one question,, how can we perform Durbin Wu Hausman test for panel data? I have found some advices on the Internet that they suggest to use Xtivreg2 and put a endog( ) option there. But if I use xtivreg2: it does not allow me to add I.year for time fixed effect. Do you know how to fix this issue? Thank you very much.

[@user-qn2nd6lm4x]

Hi, you can try this code: "xi: xtivreg2 varlist i.year, fe"

[@meganelijzen1667]

Hey! Thank you for this video, it is very helpful. I still have a question, is it possible to do a ivregression 2sls with 4 endogeneous variables and just one instrumental variable. I want to use one instrumental variable for all my endogeneous variables. I hope you can help me out. Thanks :)

[@mikejonaseconometrics1886]

In order for an IV model to be identified (and unique coefficients estimated), you will need at least one unique, external instrument for each endogenous variable in the system.

[@Dr_Shiny]

Hi Mike, For your variable "wks", you just assumed that it's endogenous, would you please share the Stata code to test the endogeneity for a variable in panel data.

[@mikemax8472]

Hi Mike, could you please tell me how to report first stage data. I used ',first' after fe to generate first stage regressions. But I don't know how to report them. Or could you tell me a journal article that reports first stage results with F tests.

[@reynaldosenrahodelin4611]

Dear Professor Jonas. I only have a question. Can we use xtivreg when all the variables are non stationary (they have unit roots)?

[@DiegoMartinez-mt2zg]

Hi mike. Thanks a lot. I have a question. How can i make a IV-FE with time fixed effects and the standard error clustered by state. the data set. I have 24 states for 5 years. i put the following command: xtivreg2 dA (Nd1 = Iv3), fe cluster(dpto). But i just obtain state fixed effect with error cluster by state o time fixed effect with error cluster by the same variables. How can i do that?

[@juandavidpenaranda6136]

Hi. How do you estimate with a time invariant variable such as the dummy fem, fixed effects IV? I have that as homework. What should I do? thanks in advance!!

[@alialipour7462]

Hi. Thanks for the video. How about HLM? I have my main analyses with stata mixed command. Doing robustness check with instrumental variables, should I use xtivreg? Or are there any other commands particular for HLM and nested data???

[@boli4359]

Hi Prof. Jonas, Would you please recommend a top journal paper with panel data instrumental variable? I learnt a lot from your video. But I guess from course to paper, there is a huge gap in application designs, identification discussion, heterogenous effects, and robustness tests. There are more to learn. Thank you very much! I hope you enjoy your research and teaching! Best Regards, Bo

[@asifnawaz5335]

Hello Mike, Can you do something on 'estadd' for regression output, specially for GMM estimations and required results to put in a table

[@Aleemmuhdr]

Hello Mike.. can you guide me on how to perform the "Wooldridge Test for Strict Exogeneity"..Thanks

[@abdullahalhejeilan1035]

Thanks Mike, this is very helpful

[@alexanderopoku6835]

i really need help on this if anybody can help. an using a panel data (two waves).

[@badiahahmed2085]

thank you for sharing this video, dear how can I interpret Hausman test result to compare between FE and FE-IV model . hausman iv ---- Coefficients ---- | (b) (B) (b-B) sqrt(diag(V_b-V_B)) | iv . Difference S.E. -------------+---------------------------------------------------------------- IHS_fdx | -10.2609 -3.488869 -6.772031 1.233345 IHS_gov | -3.616735 -3.877065 .2603298 .2091792 IHS_gfcf | 3.277582 2.25975 1.017832 .1942741 IHS_trd | 1.358966 .7292542 .6297113 .1364145 IHS_gnci | .6256677 .3832061 .2424616 .0834463 IHS_lbor | -5.572129 -4.614434 -.9576948 .5514027 dum_fdx | 3.960901 -.3721835 4.333084 1.10512 ------------------------------------------------------------------------------ b = consistent under Ho and Ha; obtained from xtivreg B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic chi2(7) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 42.89 Prob>chi2 = 0.0000 According to P-value, Which one is appropriate? THANK YOU