The coefficient on the lagged income term in technological growth models

Estimations of the growth model

income growth per person = constant0
+ constant1*income last period
+ constant2*capital accumulation
+ constant3*education increase
+ constant4*technological transfer from abroad
+ error

typically find constant1 less than zero if constant4 is set to zero before the estimation begins. So as income increases, growth declines.

If technological transfers are allowed and constant4 is estimated rather than set to zero, then my recent estimations have found constant1 to be very small, and may be positive or negative. It would seem that technological transfers account for much of the reason why growth is estimated to be lower in rich countries than in poor countries, when cross-sectional or panel data estimation is used. It also helps to explain why time series for economically leading individual countries often do not show such marked declines in their growth as would be expected from estimations.

The results are sensitive to the estimation method, and so are quite tentative at the moment.