Knowing how climate change affects the population dynamics of insect pests is critical for the future of integrated pest management. Rising winter temperatures from global warming can drive increases in outbreaks of some agricultural pests. In contrast, here we propose an alternative hypothesis that both extremely cold and warm winters can mismatch the timing between the eclosion of overwintering pests and the flowering of key host plants. As host plants normally need higher effective cumulative temperatures for flowering than insects need for eclosion, changes in flowering time will be less dramatic than changes in eclosion time, leading to a mismatch of phenology on either side of the optimal winter temperature. We term this the “seesaw effect.” Using a long‐term dataset of the Old World cotton bollworm Helicoverpa armigera (Hübner) (Lepidoptera: Noctuidae) in northern China, we tested this seesaw hypothesis by running a generalized additive model for the effects of the third generation moth in the preceding year, the winter air temperature, the number of winter days below a critical temperature and cumulative precipitation during winter on the demography of the overwintering moth. Results confirmed the existence of the seesaw effect of winter temperature change on overwintering populations. Pest management should therefore consider the indirect effect of changing crop phenology (whether due to greenhouse cultivation or to climate change) on pest outbreaks. As arthropods from mid‐ and high latitudes are actually living in a cooler thermal environment than their physiological optimum in contrast to species from lower latitudes, the effects of rising winter temperatures on the population dynamics of arthropods in the different latitudinal zones should be considered separately. The seesaw effect makes it more difficult to predict the average long‐term population dynamics of insect pests at high latitudes due to the potential sharp changes in annual growth rates from fluctuating minimum winter temperatures.