Why is that a positively expected variable happen to show a negative coefficient in panel data?

For example GDP per capita is expected to affect positively export diversification. But I keep obtaining negative and significant coefficients in fixed and random effect models.

24 August 2020 9,081 14 View