Stats question

Having something weird going on with OLS regressions. I can't disclose what the data is regarding, but here's the scenario.

If I run the regression "regress depVar on v1 v2 v3 vOfInterest v4 ..." (depVar = dependent variable, v1...vn = independent vars), I get a coefficient on vOfInterest of like 0.4. If I run the same regression on ten mutually exclusive and collectively exhaustive subsets of the data, I get coefficients for vOfInterest that all exceed 0.4. No observations or independent variables are being dropped when I do this, and there is no correlation between the dependent and independent variables. A weighted average of the coefficients with the subsets is approximately 0.79, yet the combined regression still yields 0.4. Does anyone know what else can cause this?