One of the important areas of research in Statistical Inference, namely, the selection and ranking of populations is well attended by many researchers. Here, a selection procedure is developed to select the population possessing stronger increasing failure rate average (IFRA) property among the several IFRA populations. A selection procedure is developed using a measure of departure from exponentiality towards IFRA. The strength of IFRA property of populations is determined by considering the measure given by Keerthi and Pandit [8]. The statistic proposed by Keerthi and Pandit [8] to test exponentiality against IFRA alternatives is based on a U-statistic with a kernel of degree four. Keerthi and Pandit [8] established the large sample properties of the statistic which is the basis for selection procedure. Also the performance of the selection procedure is evaluated using probability of correct selection.