摘要：O presente artigo analisa o grau de realismo do or.amento anual para 2800 munic¨Şpios, e rejeita a hip¨®tese de que a perda de realismo decorre de restriçăo informacional para estimaçăo, e encontra o uso de or.amento incremental nas despesas. A gest.o de finan.as p¨˛blicas (Performance Financial Management - PFM) tem recebido atençăo de ag¨şncias multilaterais e ¨¦ vista como base das reformas sugeridas para pa¨Şses em desenvolvimento. Em geral as intervençőes das ag¨şncias se restringem ao governo central, e n.o alcan.am a pr¨˘tica em governos subnacionais. Os resultados da an¨˘lise mostram que as reformas propostas no Brasil ainda n.o alcan.aram os resultados desejados, e que a rejei..o das hip¨®teses mencionadas sugere que nos governos locais o poder executivo antecipa oportunidades e desafios na execuçăo do or.amento, e prev¨ş brechas nas estimativas (super e sub avaliaçăo). O impacto desta pr¨˘tica ¨¦ a redu..o de accountability do processo como um todo.
其他摘要：This article analyses the degree of realism in the annual b udgets of 2800 municipalities; it rejects the hypothesis that the loss of realism results from restrictions on information for estimates, and encounters the use of incremental budgeting in expenses. The management of public finances (Performance Financial Management ¨C PFM) has received attention from multilateral agencies, and is seen as the basis of suggested reforms for developing countries. In general, interventions by such agencies are limited to national governments and do not affect the practices of local or regional governments. The results of the analysis show that the prop osed reforms in Brazil have still not brought the desired outcomes, and that the rejection of the hypotheses mentioned suggests that the executive branch anticipates both challenges and opportunities to arise during the execution of local government budgets, as well as foreseeing budgetary breaches (both over and underestimates). The impact of this practice is a reduction in accountability for the process as a whole
