CoMSES Net Discourse Forums

PolicySpace: an open-source, spatial-economic agent-based model


PolicySpace is based on concepts of urban economics, spatial heterogeneity, functional urban regions, heterogeneity of taxes collection, equal access to opportunities and the need of unified planning. In short, PolicySpace is an open-source, spatial-economic agent-based model with three markets and a tax system that empirically simulates 46 metropolitan regions in Brazil. Contributions include the explicit intra-urban space, mobility of families (via real estate Market), population dynamics (demographics), use of distance as a mediator. The model runs automated sensitivity analysis, alternative tax distributions and repeated times for each or all metropolitan regions. The model is validated in a five-step cumulative process.

Model URL:

About CoMSES 2018

Veru interesting model. How do you deal with the informal economy? In many countries, Brazil not excluded, there are illegal settlements and a lot of complex political dynamics to provide those areas services.


Interesting model. @furtadobb: could you give some more information on how Municipal Participation Funds are allocated? From what I can tell from your model those are given as CSVs. What factors are taken into account?


From what I tell from Appendix A funding is (stepwise) proportional to population but when population reaches a maximum level no additional shares are given (so smaller communities receive larger amounts of funding per person). Is that interpretation correct?


Hi. Thanks for your interest. The model does not handles informal markets specifically. However, as all the initial data is read from census data, the inherent heterogeneous of Brazilian households is embedded into the model.


Municipal funds distribution follows observed proportions until 2016 and then is the same until the end of the simulation for the default (as it is today) case. For the cases that FPM is False, then the treasure collected by municipalities are not distributed via FPM. Basically, one of the two tests is: presence or absence of FPM as a distribution rule. We find that FPM is very progressive and using FPM improves quality of life overall. In real rules for FPM, besides population there is also an ad hoc system of classes, such that municipalities above and below given thresholds have fixed coefficients. As the rule would be hard to implement, we just follow observed proportions (for the default case).


Yes. Indeed. That’s correct.


What factors in the simulation are responsible for the increase in the quality of life index from municipal aggregation? Is it because the surrounding communities in larger cities already have a higher quality of life so giving them a large of portion of funding (because they are ‘small’) has negligible quality of life index improvement? (I know in Canada people living in newer bedroom communities on the fringe of a larger city often have a higher standard of living).


The increase of quality of life comes from the investment of collected taxes by the municipalities weighted by the population. So, the more taxes collected (or received as transfers), the more QL increases (for the same given number of inhabitants). However, house prices also increase, making it harder for people to move in. So, it’s a dual feedback. This is detailed in the book. Please, check