i have 3 rounds of cross sectional survey data of firms. i wish to estimate the productivity of the firms over the years. Since the 3 surveys have different firms in each cross sections of surveys, i have to resort to create pseudo panels.

i m unable to create pseudo panels using cohorts by implementing 'collapse' command.

i realize the problem when i try to estimate the TFP using 'prodest' command, to find out that in my new data set (post 'collapse' command) no. of observations & no. of groups are same. Thus even 'prodest' command is unable to display its estimates.

Kindly refer below-

. ************************************************** ************************************************** **************************************
. cd "G:\My Drive\Adv Econometrics- lokesh FPM1808\PROJECT \output"
G:\My Drive\Adv Econometrics- lokesh FPM1808\PROJECT \output

.
.
. use data

.
. ***************************
. collapse ENTID GVA_log labour_log Loan_Tot investment_log Capital_log _GVA _TotAssetsOwned _TotAssetsHired FactorIncome_Tot
> State_code, by(Year State_Name Sector Nature_of_Op ownership FirmLocation _TotWkers)

.
.
. tabulate Year

Year | Freq. Percent Cum.
------------+-----------------------------------
2000 | 81 57.04 57.04
2011 | 34 23.94 80.99
2016 | 27 19.01 100.00
------------+-----------------------------------
Total | 142 100.00

.
.
. duplicates report ENTID Year

Duplicates in terms of ENTID Year

--------------------------------------
copies | observations surplus
----------+---------------------------
1 | 138 0
2 | 4 2
--------------------------------------

.
. duplicates drop ENTID Year, force

Duplicates in terms of ENTID Year

(2 observations deleted)

.
. egen wave=group(Year)

.
. xtset ENTID wave
panel variable: ENTID (weakly balanced)
time variable: wave, 1 to 3
delta: 1 unit

.
.
. ******Productivity Estimation - Olley and Pakes (1996)*****************************
. prodest GVA_log, method(op) free(labour_log) proxy(investment_log) state(Capital_log) id(ENTID) t(wave) valueadded
.....


op productivity estimator Cobb-Douglas PF

Dependent variable: value added Number of obs = 140
Group variable (id): ENTID Number of groups = 140
Time variable (t): wave
Obs per group: min = 1
avg = 1.0
max = 1

------------------------------------------------------------------------------
GVA_log | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
labour_log | .6997251 .0251794 27.79 0.000 .6503744 .7490759
Capital_log | .2894271 . . . . .
------------------------------------------------------------------------------
Wald test on Constant returns to scale: Chi2 = 0.19
p = (0.66)

.
. predict TFP

.
end of do-file
************************************************** ************************************************** *********************
************************************************** ************************************************** *********************