0

For an agent-based simulation, I need to model the change in the income of an individual over time.

I just got an answer about Probability distribution of income , so I can determine the income of individuals at t=0, but how would their income change over time?

Is this is a Markov series, and if so, what would be reasonable parameters for that model?

Erel Segal-Halevi
  • 791
  • 2
  • 6
  • 19
  • This is a very general question. I would suggest looking at the economic literature as a starting point rather than thinking of it as a purely statistical problem. – Peter Ellis Jul 11 '12 at 21:18

1 Answers1

3

There should be three components to the change over time:

  1. Change in the distribution of income (due to growth+inflation)
  2. Change in the conditional distribution of income (as the agent ages)
  3. Change in the agent's percentile (due to economic mobility)

The first could be extracted from the combination of total income growth (http://research.stlouisfed.org/fred2/categories/110) & income distribution (http://research.stlouisfed.org/fred2/categories/33001).

The second is difficult, as I have not been able to find a good source for age-conditional income distribution.

I don't want to endorse any specific study on economic mobility by linking here, as there are many muddled & conflicting conclusions. Intuitively, I would expect there to be greater mobility near the median than at either extreme.

Ghillie Dhu
  • 226
  • 1
  • 7
  • Thank you! Actually I was thinking mainly about #3 - the mobility. It is true that different countries have different mobility rates, but isn't there a simple parametrized model that is commonly used? – Erel Segal-Halevi Jul 12 '12 at 06:43
  • 1
    The international comparisons I've seen examined intergenerational mobility (dubiously, having failed to control for parental & child ages at sampling time). I think for an agent-based simulation you'd want intragenerational, which I haven't come across. – Ghillie Dhu Jul 12 '12 at 07:04