With 3 response categories, there are two cumulative probabilities to model, and hence 2 intercept parameters. The proportional odds form of model has the same predictor effects for each cumulative probability (i.e., the curves have the same shape), so only one effect is reported for income.