We should always check whether the residuals are homoscedastic, whether they lack a structure, and whether they are Gaussian. In this case, there seems to be an increasing variance, and that makes us believe that the assumption of homoscedasticity is not satisfactory. The normality assumption seems plausible, a time series containing the amount of dollars sold to the market by the agricultural-export sector. though . Sometimes these problems (non-normality or heteroscedasticity) can be fixed by applying logarithsm the series:

plot.ts(results$residuals) 

The following output shows the residuals:

Then, type the following code:

qqnorm(results$residuals) 

The following output shows the quantile-quantile plot. The residuals seem to be normal: