To show that the traditional econometric approach is not able to deal with deterministic chaos, I use an extension of Goodwin's growth cycle model to generate artificial data for output. An EGARCH model is estimated to describe the data generation process. Although, using some traditional econometric tests, no evidence of misspecification is found, the estimated process is qualitatively wrong: it is dynamically stable when the true process is unstable. A specific econometric procedure developed to deal with deterministic chaos is presented: the BDS statistics. Also an explanation for the little evidence of deterministic chaos in aggregated macroeconomic time series is suggested.