Objectives: Research identifying reliable and country-specific predictors of smoking is needed in order to develop effective adolescent smoking prevention programmes. The objective of this study was to assess the cognitive and socio-demographic factors associated with smoking onset among Romanian teenagers, using both cross-sectional and longitudinal data. Methods: The data were obtained from a two-wave, one-year longitudinal study carried out among 316 senior high school non-smokers from Cluj-Napoca, Romania. Questionnaires assessed smoking behaviour, attitudes, social influence, self-efficacy and intention regarding smoking (motivational variables) as well as different sociodemographic features. Results: The cross-sectional analyses showed that socio-demographics and motivational variables were strongly associated with smoking behaviour; the explained variance was 76%. The longitudinal analyses revealed that four variables explained 33% of the variance in change of status from non-smoking to regular smoking over a period of one year. Regular smoking onset after one year was predicted by baseline low self-efficacy in refraining from smoking in different situations, having more smoking friends and playing truant from school. Having a brother was a protective factor. Conclusion: The results suggest that smoking prevention programmes in Romania should strengthen self-efficacy beliefs and resistance against peer modelling and help Romanian young people to develop skills and action plans to cope with pressure to smoke and challenging situations.