“…Then, if longitudinal data were analyzed with cross-sectional models, Kmenta (1971 , p. 283) demonstrated that the residuals will be autocorrelated, the parameters ( b 0 , b 1 ,…) are not biased, but the variances of the errors are underestimated. Therefore the variances and the standard errors of the parameters (that are in the denominator) also tend to be underestimated and, likewise, the values of the t , z , F , R 2 , and b 0 , b 1 … statistics are overestimated and not efficient, leading to type I errors (the assumption that a statistical effect exists, when in fact it does not) ( Gujarati and Porter, 2013 ; Rosel et al, 2019 ). In addition, if we omit the values of the lagged variable, and this variable is part of the explanatory model of behavior, the coefficients obtained are biased and inconsistent, so the inferences drawn no longer have a substantive meaning ( Gujarati and Porter, 2013 ; Draper and Smith, 2014 ).…”