Testing additive separability

We develop a revealed preference procedure to test if violations of additivity are due to measurement error. It detects additivity for simulated data and has power to reject nonseparable data. For interwar data, additivity is detected for two “super” groups.