Detect the Change Points in the Growth Rate of US Real Export Data Based on Mean and Variance

When analyzing time series data, it is common to `ssume the data was generated form a consistent distribution or process. However, many global epidemic and financial problems caused great shake on the global economy. The shake may change the environment of US exports and thus determining the change points of statistic properties in the time series data of US exports is important. This article transforms the original US export data into the growth rate of US exports and provides researches on detecting the multiple change points in the growth rate of US real export data from 1947 to 2022. This article detects change points from the aspects of mean, variance and mean-variance under the normal distribution assumption. The algorithms utilized to identify change points in this article are the segment neighborhood and the PELT. This article verifies that the segment neighborhood can give the exact results based on fewer assumptions than the PELT. The mean-variance method gives the best results, finding 9 change points. The 9 change points corresponds to 7 segments of the entire data and the segments are coincided with the recession periods from the National Bureau of Economic Research.