Objective To study the predictive value of depressive symptoms (DeprSs) in a general population of Turkey for type 2 diabetes. Methods Responses to three questions served to assess the sense of depression. Cox regression analyses were used regarding risk estimates for incident diabetes, after exclusion of prevalent cases of diabetes. Mean follow‐up consisted of 5.15 (±1.4) years. Results Depressive symptoms were present at baseline in 16.2% of the whole study sample, threefold in women than men. Reduced physical activity grade was the only significant covariate at baseline in men, while younger age and lower blood pressure were significantly different in women compared with those without DeprS. In men, presence of DeprS predicted incident diabetes at a significant 2.58‐fold relative risk (95% confidence interval 1.03; 6.44), after adjustment for age, systolic blood pressure, and antidepressant drug usage. When further covariates were added, waist circumference remained the only significant predictor, while DepS was attenuated to a relative risk of 2.12 (95% confidence interval 0.83; 5.40). DeprS was not associated with diabetes in women, whereas antidepressant drug usage only tended to be positively associated. Conclusion Gender difference existed in the relationship between DeprS and incident diabetes. DeprS predicted subsequent development of diabetes in men alone, not in women. Copyright © 2016 John Wiley & Sons, Ltd.