Assigning causal interpretation to associations obtained from observational data is challenging as they are prone to confounding. Number of lambs born (prolificacy) in dairy sheep may be considered a potential factor contributing to milk yield (MY). However, inferring this effect using traditional regression or ANOVA techniques can generate spurious results if confounder variables affect both the outcome (MY) and treatment (prolificacy). Propensity score (PS) methods tackle this issue by balancing baseline covariate distributions between treatment levels, allowing unbiased inference of marginal effects. This method belongs to the framework of causal models dealing with potential outcomes. It intends to mimic aspects of randomized trials, in which the difference among treatment groups is causally meaningful. Under the assumption that ewe prolificacy affects MY, our objective was to estimate the magnitude of such a causal effect using PS based on Matched Samples. Data comprised 4,319 records from 1,534 crossbred dairy ewes. The set of potential confounders was composed by lactation number (1th, 2nd and 3th – 6th) and dairy breed composition (<.5, 0.5-.75 and > 0.75 of East Friesian or Lacaune). For the treatment variable, single lamb birth was assigned to Group 0, while multiple birth (2, 3 or 4 lambs) was assigned to Group 1. MY represented the volume of milk produced for the whole lactation (mean = 268.5 L and SD = 116.4 L). The R package “nonrandom” was used. A total of 1,166 pairs of treated/nontreated individuals with similar PS values were formed. The criterion for similarity was defined by a caliper size equal to 20% of the sd in the PS logit (0.13) and a ratio of treated/untreated = 1. All covariates were deemed balanced after matching (cutoff for standardized bias = 0.2). The estimated causal effect of prolificacy on MY was 20.52 L, se = 3.77 L, 95% CI = 13.13–27.91 L. This means that ewes that gave birth to a single lamb would be expected to have MY increased by 20.52 L if they had given birth to multiple lambs and all other variables were held constant. This implies that any management practice that increases (decreases) prolificacy would affect MY positively (negatively).