This paper analyzes the relationship between provincial incidence of poverty, innovation potential and economic growth, and employs difference GMM to estimate it based on the control of heterogeneity and endogeneity across regions and over time. It arrives at the conclusions that the incidence of poverty in a province has the significantly negative effect on regional innovation potential and such innovation potential indexes as human capital, central & local transfer payments of science & education and the degree of marketization are positively correlated with GDP per capita in a province. Evidence shows that the incidence of poverty in a province has the indirect effect on GDP per capita by affecting innovation potential indexes, but there is a direct negative relationship between the incidence of poverty and employment rate. In a long run, as for innovation potential indexes, the changes in the incidence of poverty have the greatest effect on human capital, and resulting role in the economy is the most serious; and the changes in the incidence of poverty have weaker effect on central and local fiscal input and resulting indirect role in the economy, but there are still small differences as for the effects. It points out that governments should pay attention to the effect of the shortage of regional innovation potential resulting from poverty on stable economic growth.