This article studies the inference procedures used to compute the macroeconomic indicators feeding into the International Monetary Fund’s monitoring and surveillance in Africa since the structural adjustment. In 2005, the IMF launched a procedure to denounce a Mauritanian “misreporting” over a twelve-year period. The article wonders how could the statistical fiction be validated by the IMF economists, and to what extent they took part in Mauritanian data production. The article argues that the auditor-auditee relation places less importance on the veracity and the pertinence of numbers than on the formal conformity of data and economic programs with expectations of the IMF bureaucracy. For programs and statistics to be considered consistent, tables of estimates must be filled out, even when data are missing, and the economic diagnosis must comply with monetarist-dominant orientations. By analyzing the financial programming tool, the article shows that the treatment of basic national accounting identities neglects variables like household consumption even when alternative methods exist. Changes in methods may be discarded to ensure the legibility of economic works over the years. The article therefore argues that the IMF and countries coproduce false accounts, whereby inferences of macroeconomic estimates serve other institutional functions inside the IMF besides veracity.