Haze is a serious issue in China with increasing concerns, and understanding the factors driving decadal-scale variations in haze occurrence is relevant for government policymaking. Using a comprehensive observational haze dataset, we demonstrate notable decadal fluctuations in the number of haze days (HD) during winter in central eastern China, showing a decline since the mid-1980s. The leading mode of the wintertime HD features an increasing trend for 1959-2012 in eastern China, highly correlated with China's gross domestic product (GDP) that represents increasing trend of pollutant emissions, and to a lesser extent meteorological factors. The second mode shows decadal variations in central eastern China associated with Pacific Decadal Oscillation (PDO). Observations and numerical simulations suggest that Mongolia High and corresponding descending motion tend to be enhanced (weakened) in central eastern China during the positive (negative) phase of PDO. With PDO shifting towards a negative phase, the weakened Mongolia High and ascending anomalies make the air unstable and conduce to the spread of pollutants, leading to the decline in the wintertime HD over central eastern China since the mid-1980s. Based on above physical mechanisms, a linear model based on PDO and GDP metrics provided a good fit to the observed HD.