In the development of our understanding of PB as a successful democratic innovation, first steps towards examining multiple cases came in the form of early comparisons of carefully selected cases of PB. While studies of individual cases in Brazil and beyond continue to dominate the literature, cross-case comparative analysis began to contribute to our knowledge. Valuable work now led by Paolo Spada (2017), building on initial data collections from Ana Clara Torres Ribeiro and Grazia de Grazia (2003), and Leonardo Avritzer and Brian Wampler (2005), has collected a census of PB in Brazil. Data begins in 1989, with a new wave of data collection in swing as I write. It shows that if survival of PB is a good indicator of its success, then the data is sobering – on average half the processes are discontinued within four years of inception (Spada and Ryan, 2017: 772). But that is getting ahead of the story. PB has now accumulated several good in-depth small-N case comparisons.

William Nylen (2003b) was early to identify the need for research to attend to ‘failed’ cases. He provides an in-depth account and compares two cases in the state of Minas Gerais – Betim and João Monlevade – where the PT was voted out of office following a clear opposition mandate to remove all traces of such participatory programmes (2003b: 96). Nylen contrasts the overt mobilization of petistas (affiliates or supporters of PT) at the expense of working with existing neighbourhood associations in João Monlevade's first shortlived PT administration, with subsequent conscious efforts to avoid partisan mobilization in Betim's PB.

Although the PT survived an extra term in Betim, the result – an end to PT rule and to participatory programmes in the face of fierce political competition – was not averted. In both cases Nylen reports that a strong and sometimes violent opposition were able to recast PB as partisan and inefficient regardless of the PT's efforts. Despite the variance in effort, the result was the same. Based on Mill's principle of agreement we can discard these active efforts to avoid partisanship by a ruling party instigating PB as a sufficient condition for enduring change. An all too often ignored strength of these simple comparative techniques in the empirical social sciences is the potential to discount factors that might be theorized to be relevant to an explanation.