This study measures hypothetical bias in CVM-based estimates of willingness to support a forest protection and restoration program in Poland. It does so in two parallel ways: within- and between-subjects. To this end, participants are asked to vote in a referendum on an NGO-led reforestation project for various per-capita cost levels. They do so either directly or, as a surprise continuation, after a hypothetical analogue. Both methods deliver a similar (and substantial) level of hypothetical bias – it is slightly higher, but not significantly so, in the between-subject comparison. This suggests that hypothetical bias in this context is not primarily driven by such factors as consistency seeking or impression management.