This paper examines the time series persistence of economic policy uncertainty in China using a newly developed time-varying parameter autoregressive framework. We apply this methodology to two distinct news-based measures: one derived from mainland Chinese newspapers and another from Hong Kong's South China Morning Post. The analysis reveals a key divergence: the mainland index exhibits lower, more volatile persistence, while the SCMP-based measure shows higher and more stable persistence. This finding suggests that news-based uncertainty measures are not neutral - they reflect the institutional context that produces them. The results have implications for how we measure uncertainty, and for understanding how media bias shapes not just the content of information, but its temporal dynamics. Thus, state media control shapes information in measurable ways.