We develop a time-varying measure of cay (cay TVP) using time-varying cointegration, and then compare the predictive ability of cay TVP with cay and a Markov-switching cay (cay MS) for excess stock returns and volatility in the US over the period 1952:Q2-2015:Q3, using a kth order nonparametric causality-in-quantiles test. We find that time-varying cointegration exists between consumption, asset wealth, and labor income. In addition, while there is no evidence of predictability of volatility of excess returns from cay, cay MS , or cay TVP , they tend to act as strong predictors of stock returns, with cay TVP being important during the bearish phases of the equity market.