The use of instruments originally developed for measuring gambling activity in younger populations may not be appropriate in older age individuals. The aim of this study was to examine the presence of problematic and disordered gambling in seniors aged 50 or over, and study the reliability and validity properties of the SOGS (a screening measure to identify gambling related problems). Two independent samples were recruited: a clinical group of n = 47 patients seeking treatment at a Pathological Gambling Outpatient Unit, and a population-based group of n = 361 participants recruited from the same geographical area. Confirmatory factor analysis verified the bifactor structure for the SOGS with two correlated underlying dimensions [measuring the impact of gambling on the self primarily (Cronbach's alpha α = 0.87) or on both the self and others also (α = 0.82)], and a global dimension of gambling severity (also with excellent internal consistency, α = 0.90). The SOG obtained excellent accuracy/validity for identifying gambling severity based on the DSM-5 criteria (area under the ROC curve AUC = 0.97 for discriminating disordered gambling and AUC = 0.91 for discriminating problem gambling), and good convergent validity with external measures of gambling (Pearson's correlation R = 0.91 with the total number of DSM-5 criteria for gambling disorder, and R = 0.55 with the debts accumulated due to gambling) and psychopathology (R = 0.50, 0.43 and 0.44 with the SCL-90R depression, anxiety and GSI scales). The optimal cutoff point for identifying gambling disorder was 4 (sensitivity Se = 92.3% and specificity Sp = 98.6%) and 2 for identifying problem gambling (Se = 78.8% and Sp = 96.7%). This study provides empirical support for the reliability and validity of the SOGS for assessing problem gambling in elders, and identifies two specific factors that could help both research and clinical decision-making, based on the severity and consequences of the gambling activity.