OBJECTIVES: Human papillomavirus (HPV) vaccines prevent infection with the most cancer-causing HPV types. However, many age-eligible girls do not initiate or complete the three-dose regimen. There is limited information on the magnitude of protection provided by partial-vaccination. We sought to empirically estimate the effect of HPV vaccination by dose (0, 1, 2 or 3) on cervical abnormality rates in a screening cohort of privately-insured U.S. girls (N=234,829).  METHODS: We defined a cohort within the MarketScan Commercial Claims database that included 9-17 year-olds in 2007 (year of vaccine introduction) who began Pap screening during 2008-2013. Follow-up started at subjects’ first-ever Pap and ended at abnormality detection or censoring. Treatment group corresponded to doses received before the first screening. We used inverse probability of treatment weighting to adjust for birth year, age/year of first Pap and first dose, plan type, region, zipcode-level socioeconomic factors, and claims-based sexual risk proxies. Weighted pooled logistic regressions estimated the effect of dose level on abnormality rates. RESULTS: Subjects received zero (63%), one (8%), two (9%), or three (20%) doses (mean ages at first Pap and first dose: 17.8 and 15.6 years). Adjusted incidences of any cytological abnormality (events per 100 annual person-visits) ranged between 11.61 with zero doses and 9.10 with three doses. Incidence rate ratios (IRRs) versus zero doses were 0.91 (95% CI: 0.87-0.96), 0.83 (95% CI: 0.80-0.87), and 0.76 (95% CI: 0.74-0.79) for one, two, and three doses, respectively. For histology-confirmed high-grade lesions (CIN2+), IRRs versus zero doses were 0.63 (95% CI: 0.51-0.77), 0.68 (95% CI: 0.58-0.82), and 0.47 (95% CI: 0.40-0.55) with one, two, and three doses, respectively.  CONCLUSIONS: HPV vaccination may substantially reduce the risk of cervical abnormalities in real-world settings, even when vaccination is incomplete. Policy and clinical efforts should focus on encouraging eligible girls to initiate vaccination.