Can Measurement Errors or Seasonal Adjustments Explain the Negative Autocorrelation of U.S. Monthly Consumption Changes?

The negative first-lag autocorrelation of U.S. monthly consumption changes rejects the continuous-time random walk model of consumption. This paper addresses the question of whether data distortions due to measurement errors or the application of the X-11 filter may explain this negative autocorrelation and thus possibly rescue the model from failure. Regarding measurement errors, the paper finds that the conditions under which plausible types of errors may explain it violate a cross-frequency restriction. Regarding seasonal adjustment, the paper finds that if the unknown seasonal component in consumption is stochastic with unit seasonal roots, the application fo the X-11 filter makes the first-lag autocorrelation positive instead of negative; if the unknown seasonal component is stationary, it can make it positive or negative depending on the unknown parameters of the seasonal process. Overall, though far from proving an impossibility theorem, the paper provides some arguments against the continuous-time random walk model.