This study investigates the impact of an error correction method for a Japanese handwritten answer recognition system on the accuracy of automated grading. Previous research has demonstrated the effectiveness of T5-based error correction for automated transcription of handwritten Japanese answers; however, its influence on automated grading systems has not been thoroughly investigated. In this study, we applied corrected recognition results to a zero-shot automated grading system using ChatGPT and compared grading accuracy before and after error correction. Experimental results showed that error correction improved grading accuracy by up to 13% points. Furthermore, we conducted a detailed analysis of grading cases to elucidate the ways in which error correction influences grading outcomes. These findings underscore the potential of error correction methods to enhance the accuracy of automated grading for descriptive answers.