Purpose: The paper aims to prove that between 1992 and 2016, people in poverty as a proportion of the total population has not been reduced. In particular, food poverty (FP) represented an average of 22%, despite the fact that gross domestic product (GDP) per capita and GDP, social development expenditure and food programme expenditure (both as GDP proportion) grew by 0.96%, 1.9%, 2.7% and −17.4% on an annual average, respectively. Design/methodology/approach: There are non-linear relationships between economic growth and food poverty expenditure to reduce poverty. Three econometric models were estimated as follows: a linear model [ordinary least squares (OLS)] that addresses the capability of the economic growth to reduce FP (which detects a structural change in 2007) and three models of regime change (Markov–Switching Regression) that prove the existence of two different regimes. Findings: The author proved that economic growth has lost the capability to reduce poverty and that there are decreasing effects of expenditure in addressing poverty since 2007. These results point out that Mexico is in a poverty trap and suggests that for the economy as for life and even more in the case of social (public) policies, more is not necessarily better than less. Therefore, the author suggests that the resources allocated in response to poverty may well have generated perverse incentives that yielded the opposite results. Originality/value: There is no official measure of the public expenditure for poverty. Therefore, an accurate series was built to estimate the government effort and do the econometrics that proves the main hypothesis. This is another contribution. [ABSTRACT FROM AUTHOR]