xtabond2: How to interpret coefficient?

[Elena Pavia]

Goodmorning,

I'm analysing the effect of electronic payments on economic growth in 27 EU Countries from 2000-2016.
I want apply system GMM to my panel dataset, so I used xtabond2 command.

This is my input in STATA:
xtabond2 lg_gdp l.lg_gdp cpay ctr ddebts cash int_rate sepa, gmm(lg_gdp cpay ctr ddebts cash int_rate, collapse laglimits(2 3)) ivstyle(sepa, equation(level)) nocons twostep small artest(4)
where:
lg_gdp is the logarithm of real GDP per capita at time t
cpay is the ratio between the volume of debit and credit card transactions over real GDP
ctr is the ratio between the volume of credit transfer transactions over real GDP
ddebit is the ratio between the volume of direct debit transactions over real GDP
cash is the ratio between the volume of dcash transactions over real GDP
int_rate is a short period interest rate
sepa is a time dummy variable that takes value 0 before 2008 and value 1 after 2008, it indicates the adoption of Single Euro Area Payments among EU Countries.

All test statistics (AR, J Hansen and Differences-in-Hansen) not reject H0, so my regression seems valid.

Anyway, I can't understand how to interpret the estimated regressor coefficients, because on the left side I have a logarithm for the real GDP per capita and on the right side I have ratios for electronic payments.

I attach the STATA output in order to make more clear my question.

Thanks for any advice!

[Sebastian Kripfganz]

Holding everying else constant, a one percentage point increase in the ratio between the volume of debit and credit card transactions over real GDP yields a 0.06 percentage points increase in real GDP per capita growth (since you can subtract the lag of the log of real GDP per capita on both sides and the resulting first difference of the logs on the left-hand side approximates growth rates).

[Elena Pavia]

Mr. Kripfganz, thank you for your answer, now the intepretation is clear!

[Elena Pavia]

Goodmorning,

I have another doubt about my model and System GMM estimator in general.
If my regressors, i.e. the ratio between electronic payments and realg GDP, are affected by substitution effects because, for example, when cards over real GDP increase credit transfer over real gdp decrease, my model could suffer from this or could be invalidated by this?

Thanks for any answer!