¹⁸In practice it is common for several question to be omitted at once, so in these cases we cannot estimate separate coefficients for each missing category due to perfect multicollinearity among these categories. In this case only one coefficient is estimated capturing missing responses.