Mark and Sul OBES 2003 Panel DOLS

[TomDoan]

This is the analysis of the money demand function from Nelson C. Mark & Donggyu Sul, 2003, "Cointegration Vector Estimation by Panel DOLS and Long-run Money Demand," Oxford Bulletin of Economics and Statistics, vol. 65, no. 5, 655-680. It matches the point estimates, with somewhat different t-stats. The covariance matrix for DOLS depends upon an estimate of the LR variance of the residuals which can be computed quite a few ways; after working through the example, we found a minor error in the author's original calculations, and also switched to an estimator which is more stable in small cross-sections.

Note the use of the SWEEP instruction for cleaning out the individual specific "nuisance" terms. This is also done in the Pesaran, Shin, Smith example: http://www.estima.com/procs_perl/pesaranshinsmithjasa.zip, which has some similarities to Mark and Sul (common cointegrating vector, heterogeneous adjustments).

Program file for PDOLS:

moneydemand.rpf

(2.87 KiB) Downloaded 441 times

Program file for Single Country Estimates:

singles.rpf

(1.25 KiB) Downloaded 266 times

Data file:

panelmoney.xls

(58 KiB) Downloaded 398 times

Supporting procedures:

pdolsrobust.src

(1.25 KiB) Downloaded 394 times

amunivariate.src

(1.88 KiB) Downloaded 317 times

The single country estimates use the SWDOLS procedure (viewtopic.php?f=7&t=1052).

[flopez2011]

I have tried to replicate the Mark and Sul (2003) paper and I got only the last two rows (panel estimates), I do not how can I get the single-equations Dols estimates i.e. the countries regressions results shown in the Table 7 at the 678 page...Can someone help me?

[TomDoan]

I have tried to replicate the Mark and Sul (2003) paper and I got only the last two rows (panel estimates), I do not how can I get the single-equations Dols estimates i.e. the countries regressions results shown in the Table 7 at the 678 page...Can someone help me?

That's been added to the post above as the SINGLES.RPF program. Note that this also requires a revised version of the SWDOLS procedure.

[Poppelwops]

The program really works fine, but I have two problems, which I encountered when modifying the program for my own data set.

1.) Is it true that a country is excluded from the analysis if there are only some missing observations?

2.) Why are the last and first estimated coefficients of the time dummies always zero? (this question may actually be quite stupid).

Any help would be great! Cheers, Poppelwops

[TomDoan]

The program really works fine, but I have two problems, which I encountered when modifying the program for my own data set.

1.) Is it true that a country is excluded from the analysis if there are only some missing observations?

I don't see anything in this that requires a balanced sample.

2.) Why are the last and first estimated coefficients of the time dummies always zero? (this question may actually be quite stupid).

There's a comment in program about that:

Code: Select all

* Case two - no time trend, with common time effect. (There are some
* extra dummies in tdummies that are for periods lost in taking leads
* and lags. That has no effect on the results.) This uses the same sweep
* variables as case one.