A two-component model for the evolution of real GDP per capita in the United States is presented and tested. First component of the growth rate of GDP represents the growth trend and is inversely proportional to the attained level of real GDP per capita, with the nominator being constant through time. Second component is responsible for the fluctuations around the growth trend and is defined as a half of the growth rate of the number of 9-year-olds. This nonlinear relationship between the growth rate of real GDP per capita and the number of 9-year-olds in the US is tested for cointegration. For linearization of the problem, the population time series is predicted using the relationship. Both single year of age population time series, the measured and predicted one, are shown to be nonstationary and integrated of order 1 ï¿½ the original series have unit roots and their first differences have no unit root. The Engel-Granger procedure is applied to the difference of the measured and predicted time series and to the residuals of a linear regression. Both tests show the existence of a cointegrating relation. The Johansen test results in the cointegrating rank 1. Since the cointegrating relation between the measured and predicted number of 9-year-olds does exist, the VAR, VECM, and linear regression are used in estimation of the goodness of fit and root mean-square errors, (RMSE). The highest R2=0.95 and the lowermost RMSE is obtained in the VAR representation. The VECM provides consistent, statistically reliable, and significant estimates of the slope in the cointegrating relation. Econometrically, the tests for cointegration show that the deviations of real economic growth in the US from the growth trend, as defined by constant annual increment of real per capita GDP, are driven by the change in the number of 9-year-olds.