Clinicians use nodule size to determine which thyroid nodules should receive cytological evaluation. The American Thyroid Association (ATA) has recommended against cytological evaluation for nodules <1 cm. It is unknown, however, if nodule size can accurately discriminate lesions that will represent tumors with favorable versus unfavorable prognosis. Also, the characteristics of thyroid cancers that would not be diagnosed if a strict 1 cm size cut off is used as the threshold for biopsy of intrathyroidal nodules are not well established. Using the Rochester Epidemiology Project, a population-based cohort, we identified all thyroid nodules in Olmsted County residents from 2003–2006. To assess the presence of favorable or unfavorable features for each nodule size cutoff, each patient found to have thyroid cancer was risk-stratified using the ATA risk score, which predicts risk of recurrence and persistent disease. Thyroid cancer cases in which a biopsy was done for factors other than thyroid nodule size or suspicious ultrasound features were excluded. We identified 485 thyroid nodules, 46 (9.5 %) harbored thyroid cancer. Of the 46 thyroid cancers, 37 (7.6 %) had ATA low risk; 8 (1.6 %) had intermediate, and only 1 (0.2 %) had an ATA high risk scores. The frequency of thyroid cancer and the distribution of ATA risk scores were similar across tumor sizes. In thyroid nodules of <1 cm, 92 (87 %) were benign, while 13 (13 %) were malignant (11 % ATA low risk, 2 % ATA intermediate risk) without extrathyroid extension, aggressive histology, or distant metastasis. For all thyroid cancer patients, no cases of persistent disease were found after a median follow-up of 7 years. In this population-based study, we showed that high risk thyroid cancers are rare; indeed, in this highly selected cohort of patients, the ATA’s recommendation to avoid cytologic evaluation in thyroid nodules less than 1 cm would not miss any thyroid cancer with high risk features. However, thyroid nodule size at presentation did not accurately discriminate between tumors with favorable versus unfavorable clinicopathologic features. Thus, if further discrimination is desired, for example, to avoid overdiagnosis, features other than size at presentation need to be evaluated.