the following growth model:

ΔYit = (β0′ − 1)Yi,t−s +β1′Xi,t +β2LSi,t−s′ +μi +δ t +υit, (1)

where subscripts i = 1, . . . , N and t = 1, . . . , Ti indicate country and year, respectively,

υit ∼ IID(0,συ2 ), and uit is independent of mi and serially uncorrelated.

The above model is equivalent to the following:

Yit = β0′Yi,t−s +β1′Xi,t +β2LSi,t−s′ +μi +δ t +υit, (2)

where Yit is the log level of real per capita GDP at purchasing power parity (PPP), and DYit is the log difference in per capita GDP.As I take the first difference of the above equation, I actually subtract lagged income from both sides of the equation, giving the dependent variable as economic growth regressed on initial income and a set of control variables including labor standards. Concerning the set of controlling variables, Xit, it includes enroll for primary school enrollment, grosskform for investment, popg for the

growth rate of the population, trade for the degree of openness, and govfinalcons for the size of the government. LSit are the measures of labor standards: strike represents the rate of strike and lockouts, and injury indicates the rate of work injuries.

Table 4, I have two sets of model specifications: each one corresponds to a

different measure of labor standards: strike (specifications (1) to (3)) and injury (specifications (4) to (6)). I also construct two cumulative variables of injury and strike and present the respective estimates with specifications (3a) and (6a). If I assume that labor standards may only change a little over time, then a cumulative variable might capture its association with economic growth. II start with a baseline economic growth model that includes measures of human capital and physical capital, labor force, and one measure of labor standards.This baseline model is then augmented, one step at a time, by including the extent of international openness and the size of the government.

In the Author's table: logGDP is the dependent variable. The independent variables are: l1.loggdp l5.logGDP enroll trade grosskform governmentexp populationgrowth. The author as mentioned above introduces l.strike or l.injury.

Here is basically what we have tried:

xtabond2 loggdp2005 l1.loggdp2005 l5.loggdp2005 enroll grosskform popgrowthpercent trade l.strike if year>1997, gmm(l1.gdp2005 l5.gdp2005 l.strike, lag (3 3)) iv ( ) r small

The author mentioned the third lag in his appendix.

I can send you the article and its appendix is you need it.

Kent.

xtabond2 loggdp2005 l1.loggdp2005 l5.loggdp2005 grosskform popgrowthpercent l.strike if year >1997, iv ( l1.loggdp2005 l5.loggdp2005 grosskform popgrowthpercent l.strike l3.grosskform l3.popgrowthpercent l4.strike ) r small