LITMIR.BIZ

      1 1 Demographic data relative to 2019 (Seade 2019) and all other data relative to 2018 (infocidade.prefeitura.sp.gov.br and observasampa.prefeitura.sp.gov.br/gestao‐publica).

      2 2 www.londoncouncils.gov.uk/node/31572 and https://council.nyc.gov/budget/fy2016.

      3 3 www.saopaulo.sp.leg.br/vereadores.

      4 4 The literature on service delivery, on the other hand, presents good accounts of the conditions under which services can be improved and delivered in Southern countries. However, these studies analyze national and regional transformations, indicators and processes in countries such as South Africa (Khosa 2000; McLennan and Munslow 2009; Palmer et al. 2017), India (Vivek 2015) and Brazil (Tendler 1997). This literature has illuminated several essential processes, but cities participate merely as scales of data aggregation, even in the case of indicators of their technical capacities.

      5 5 The comparative network involves colleagues from University College London, Sciences Po, Colegio de México and the University of Milano Bicocca, with Professors Patrick Le Galès, Mike Raco, Claire Colomb, Vicente Ugalde and Alberta Andreotti, Tommaso Vitale, among others.

      6 6 It is relevant to add Holland’s (2018b) arguments refining this hypothesis for young democracies with the so‐called truncated welfare states in Latin America. The author sustains that when welfare policies mainly and continuously benefit the middle classes and the rich more than the poor, the political calculus of each of these groups happens differently. The rich, and not the poor, would favor redistribution, changing the support for welfare state formation and political parties. The argument is compelling but seems more appropriate to Brazil before the 1990s when the so‐called regulated citizenship (Santos 1979) was still central. Since then, many expansions of rights and policies have taken place, regardless of all the problems they still present (Arretche 2018). Therefore, it is the poor that recognize themselves in social policies in contemporary Brazil, as some studies have been showing (Arretche et al. 2016).

      7 7 Despite a large number of political parties in Brazil, classifying them (and governments) within an ideological continuum is not tricky. We follow widely used classifications produced from party behavior in Congress. These are also consistent with surveys among politicians about substantive positions (Figueiredo and Limongi 1999).

      8 8 This distinction can also be read using Lowi's (1972) classical definition, with all governments producing distributive policies (that apparently benefit all, and thus have low political costs), but just left‐wing governments producing redistributive policies (that clearly benefit some at the expense of others).

      9 9 Like many other processes, policy resilience is also influenced by actors and processes located at other levels, as we discuss later.

      10 10 Table I.2, at the end of the introduction documents the information, already advancing details covered by the chapters.

      11 11 To calculate the incidences cited here, we coded the information per policy and government into three different tables. All cells were initially coded 0 but were altered to 1 if that government: had initiated or reinitiated that policy considering all implementation intensities; considering only medium and robust implementation; as well as whether the program merely existed (with any intensity) during that government. For restarts from latency we considered as one both reinitiating with any intensity or jumping two points in implementation, from weak to strong. The coded information allowed the calculation of distributions by ideological blocks and the cited statistical tests.

      12 12 All these differences between the left and any other block (or combination) were significant at 95% in tests of average difference.

      13 13 Higher averages for the left continued to be significant at 95% compared to the center‐right, right or both.

      14 14 Considering all stops, additionally to reductions of intensity from strong or medium to weak implementation.

      15 15 The cross‐tabulation between trajectories and types of redistribution is significant at 95% considering several statistical tests and measures.

      16 16 Elections for state capitals returned in 1985, but elections for other cities continued without interruption during the dictatorship. Before 1985, mayors of state capitals (along with some other special situations) were appointed by state governors.