In this study, we investigate the impact of plant size on productivity of each manufacturing sector in the United States from 1914 to 1939. We measure productivity as the value of output divided by total labor, and plant size as the value of total labor divided by the number of plants. Classifying sample data by industry similarity, plant size, and time(i.e., before or after the Great Depression), we conduct three regression analyses to examine heterogeneous effects. Manufacturing-level data comes from the Statistical Abstract of the U.S. The results show that when the plant size is not large enough, its correlation with productivity is negative. However, when the plant size has reached a certain level, the coefficient becomes positive. The analyses are referable for both governments and companies wanting to promote productivity by adjusting establishment size. Developing countries in similar economic status as that of the U.S. during this historical period will also gain valuable insights on a way of flourishing national productivity.