The Newey-West procedure is meant to adjust the covariance matrix of the parameters to account for autocorrelation and heteroskedasticity. It is typically used in financial applications when one estimates the alpha (a parameter in a regression model) of a portfolio or strategy. One would adjust the standard errors using the Newey-West procedure in order to obtain a better t-statistic to determine whether the strategy generates significant returns. In the case of multivariate regression, one can calculate the covariance matrix of the parameters, e.g. the covariance matrix that measures the uncertainty in your estimates of alpha and the beta in the above example. The standard error is merely derived from this matrix. Hence, correcting this covariance matrix of the estimates, leads to changing standard errors, and different results to t-tests.
In the paper you linked to, they specifically refer to Newey-West standard errors, which is essentially the same thing as what I described above. They are worried about the impact of overlapping periods on the test statistic. This would not impact the regression coefficients, which is what seems to be your problem.