Validation

Validation is made by a succession of steps.

1. First a series of macroeconomics indicators have to behave reasonably within expected values
2. These indicators have been calibrated to be in such a way. However, they happen to be reasonably simultaneously. That means GINI coefficient + inflation numbers + unemployment among others are all comparable to observable data.
3. Rules and procedures have come from literature and or data.
4. Some of the rules can be turned off and back on, allowing users of the model to "trust" or not individual rules.

Validation itself

1. Is done by comparison data collected for Brasília real estate market that never goes into the model. It was gathered independently from estate offers freely available on the internet mostly during 2020.
2. Model data, however, only uses information from Census and official data, mostly from 2010 (although it is also possible to run with 2000s data)

Comparison

1. We compare normalized price over space for Brasília.
2. We compare hedonic regression results for simulated and real data for prices against characteristics and location.
3. We compare the histogram of prices of houses.

Results

1. We believe the model replicates the structure and the mechanisms of the real estate market.
2. However, it does not reflect specific neighborhoods. Mostly, the model predicts higher prices on neighborhoods that have firms -- and jobs offers, whereas real data suggests that entirely residencial areas are valued much more than "fundamentals" would indicate.

How to run these

1. Provide the CSV files and run individually.
2. CSV files of the simulated come from running the model
3. Real data will be provided (if not too large)