Rates of gambling and gambling-related harm fluctuate over time, influenced by availability, adaptation and demographic change, among other things. Assessing change is compromised by methodological variation. The main aim of this paper is to assess change in gambling participation and problems in adult Victorians over a 5 year period. Data are from the Victorian Gambling Study (VGS) 2008–2012 (n = 15,000) and the 2003 Victorian Longitudinal Attitudes Survey (n = 8479). An additional aim was to determine the impact of methodological differences on prevalence estimates. Despite gambling availability increasing and more activities being included participation rates declined substantially. Decreases occurred across almost all demographic groups and gambling activities. When adjustments were made for methodological differences there were no significant changes in problem, moderate risk and low risk gambling. Males and people with lower education had higher rates in both surveys. In the latter survey, two groups that experienced large participation reductions, namely young adults and metropolitan residents, emerged as additional groups with higher rates of problem and moderate-risk gambling. Further research is required to discover why overall rates of harm may have plateaued when participation continues to fall and why some groups with reduced participation experience increased harm. The findings suggest that availability and total consumption models are over-simplistic. They further suggest that to be effective prevention programmes will need to extend beyond gambling availability to include interventions directed towards individuals at risk and wider environmental determinants of vulnerability and harm. Additionally this study found that restricting administration of the problem gambling measure to subsets of gamblers generate significantly lower prevalence estimates, implying that many previous surveys under-portray gambling-related harm and that without appropriate adjustment for methodological variation findings cannot be validly compared across studies.