Distributions of permanent income and wealth in the SCF.

Micro and macroeconomic models dealing with the distribution of wealth and its accumulation often need to make assumptions about the initial distribution of wealth, ``permanent income'', or their ratio. This repository is meant to provide clear and replicable moments of these distributions that can be used by researchers in their calibrations.

Inputs

This repository's summary statistics are calculated on the Survey of Consumer Finances' Summary Extract files. For replicability, it uses the Summary Files' readily computed values of income, wealth, age, and education.

Output: WealthIncomeStats.csv

This is the main file that contains the summary statistics that are the purpose of this repository. They are stored as a .csv table with the following columns, described by groups:

• Demographic and sample-defining variables: these variables describe the sample on which the summary statistics were computed on. Their values should be read as filters applied to the SCF's population before computing the summary statistics.

  • Educ: education level. It can take the levels NoHS for individuals without a high-school diploma, HS for individuals with a high-school diploma but no college degree, and College for individuals with a college degree. All marks rows in which individuals from all educational attainment levels were used.
  • Age_grp: age group. I split the sample in 5-year brackets according to their age and this variable indicates which bracket the statistics correspond to. Note that the left extreme of brackets is not included, so (20,25] corresponds to ages {21,22,23,24,25}. All marks rows in which all age groups were used.
  • YEAR: survey wave of the SCF. It indicates which waves of the SCF were used in calculating the row's statistics. All marks rows in which all waves were pooled. NOTE: when combining multiple waves, I do not re-weight observations: I continue to use the weight variable as if all observations came from the same wave.

• Summary statistics:

  • lnPermIncome.mean and lnPermIncome.sd: survey-weighted mean and standard deviation of the natural logarithm of "permanent income", as measured by the variable norminc in the SCF's summary files. Note that this measure contains, among others, capital gains and pension-fund withdrawals and might therefore substantially differ from the popular concept of the "permanent component" of labor income, especially at older ages (see the SCF's website for exact definitions).

  • lnWealth.mean and lnWealth.sd: survey-weighted mean and standard deviation of the natural logarithm of "wealth", as measured by the variable networth in the SCF's summary files. Find the definition of networth in the SCF's website.

  • lnPermIncome.mean and lnPermIncome.sd: survey-weighted mean and standard deviation of the natural logarithm of the ratio of wealth to permanent income, as measured by
    networth/norminc.

• Weights, observation, and base year:

  • obs: counts the number of observations in from the SCF that go into computing the statistics in each row. Note that the SCF creates five "imputation replicates" for every individual that is actually surveyed. Therefore the actual number of individuals that go into computing each row should be approximately obs/5.
  • w.obs: is the sum of survey weights of the observations that are used for computing the summary statistics in each row.
  • BASE_YR: indicates the base year used for expressing dollar quantities in the SCF summary files used for computing the summary statistics. The ratio of wealth to permanent income won't be affected by inflation adjustments, but permanent income and wealth (independently) will be.

Replicating the output table

Running the script ./do_all.R will download the necessary files from the SCF's webpages and calculate the summary statistics.

Make sure to specify the base year in which nominal quantities are expressed in ./do_all.R.