Based on the analysis of urban and rural residents’ per capita disposable income and consumption levels in Chongqing from 1995 to 2018, it is found that although the regression of the mean fits the data well, there is heteroscedasticity from the residual plot.It will tend to the biased regression results for the per capita disposable income and consumption relationship when we use the least square method but the quantile regression can be a very good solution to this problem.Therefore, it was decided to analyze this data using quantile regression.The results show that median regression is better than mean regression in terms of the average absolute error of the fitting; At different quantiles, the per capita income and consumption of urban and rural residents are always positively correlated; From the coefficient estimation map of the 99 quantiles,the per capita income of urban residents at low quantiles has a greater impact on the consumption levels of urban residents while high quantiles have a smaller impact;For rural residents, on the contrary,the per capit income has a greater impact at middle and high quantiles.