Is there an inverse U-shaped relationship between pay and performance?