The Penn World Tables

The PWT’s main purpose is to construct panel data for GDP and its components in constant international prices.


National Accounts report GDP and its components in local currency units (LCU).

We could use exchange rates to make these figures into dollars.

The data would then imply that people in low income countries are shockingly poor, not just very poor.

This would be misleading because prices are systematically lower in low income countries.

To compare living standards across countries (and over time), it is necessary to deflate GDP with the local price of a consistent bundle of goods.

How Is This Done?

The ICP collects data on local prices every several years (benchmark years).

Between benchmark years, prices are interpolated.

Even though there is no theoretically “best” price index, there is a substantial theory of how to construct price indices.

The PWT does this and reports GDP in international prices.

This comes in 2 flavors:

  • RGDPE: to compare living standards (think: deflated by the consumer price index)
  • RGDPO: to compare productive capacity (think: deflated by a producer price index)

Obtaining the data

The data can be downloaded here.

I recommend downloading the Stata file and converting into a matlab dataset using Stat/Transfer.

Format of the Stata file:

  • each row is a country / year combination
  • countries are identified by their ISO codes. These are 3 letter abbreviations.
  • each column is a variable (except for the first 4 columns which are country / year info)

Since Mathworks plans to phase out datasets, it would make sense to then convert the whole thing into a Matlab table using dataset2table.

It would now make sense to make each variable into an array indexed by [country, year].

  • Fortunately, someone has already done this, so you can download the result from the PWT web site.
  • Unfortunately, that file contains errors (as of 2014-Dec). So you cannot use it!