Can survey participation alter household saving behavior?

Much empirical research in economics is based on data from household surveys. Panel surveys are particularly valuable for understanding dynamics and heterogeneity. A possible concern with panel surveys is that survey participation itself may alter subsequent behavior. We provide novel evidence of survey effects on a central life-cycle choice: household saving. We exploit randomized assignment to survey modules within the LISS Panel, an internet panel survey which is representative of the Dutch population. We find that households that respond to detailed questions on expenditures and needs in retirement reduced their non-housing saving rate by 3.5 percentage points, on average. This mean effect is driven by high-education households which have the highest pension and housing wealth. Our saving measure is based on linked administrative wealth data. Thus we can rule out the possibility that the effect is on reporting, rather than on the underlying saving behavior. One interpretation is that the survey acted as a salience shock, possibly with respect to reduced housing costs in retirement.