Survival probability is fundamental for understanding population dynamics. Methods for estimating survival probability from field data typically require marking individuals, but marking methods are not possible for arthropod species that molt their exoskeleton between life stages. We developed a novel Bayesian state‐space model to estimate arthropod larval survival probability from stage‐structured count data. We performed simulation studies to evaluate estimation bias due to detection probability, individual variation in stage duration, and study design (sampling frequency and sample size). Estimation of cumulative survival probability from oviposition to pupation was robust to potential sources of bias. Our simulations also provide guidance for designing field studies with minimal bias. We applied the model to the monarch butterfly (Danaus plexippus), a declining species in North America for which conservation programs are being implemented. We estimated cumulative survival from egg to pupation from monarch counts conducted at 18 field sites in three landcover types in Iowa, USA, and Ontario, Canada: road right‐of‐ways, natural habitats (gardens and restored meadows), and agricultural field borders. Mean predicted survival probability across all landcover types was 0.014 (95% CI: 0.004–0.024), four times lower than previously published estimates using an ad hoc estimator. Estimated survival probability ranged from 0.002 (95% CI: 7.0E−7 to 0.034) to 0.058 (95% CI: 0.013–0.113) at individual sites. Among landcover types, agricultural field borders in Ontario had the highest estimated survival probability (0.025 with 95% CI: 0.008–0.043) and natural areas had the lowest estimated survival probability (0.008 with 95% CI: 0.009–0.024). Monarch production was estimated as adults produced per milkweed stem by multiplying survival probabilities by eggs per milkweed at these sites. Monarch production ranged from 1.0 (standard deviation [SD] = 0.68) adult in Ontario natural areas in 2016 to 29.0 (SD = 10.42) adults in Ontario agricultural borders in 2015 per 6809 milkweed stems. Survival estimates are critical to monarch population modeling and habitat restoration efforts. Our model is a significant advance in estimating survival probability for monarch butterflies and can be readily adapted to other arthropod species with stage‐structured life histories.