This is the first of the transmissions of the data from my paper “Accounting for U.S. Real Exchange Rate Changes.”

As you are probably aware, there is nothing quite so simple in empirical work as “the data.” There is the raw data, the corrected data and then various stages of transforming the data. You didn’t really say which you wanted, so I figure it is probably best to send you everything.

This message and the next one contain the data for section 1 on CPIs. At the time I was working on the paper, I wrote a Word document (Index.doc) that does a fairly good job documenting the data. The CPI data is stored on my computer in two different subdirectories. The first contains, roughly, the raw data and corrected data. The second contains transformed data and data files and programs that I used for the paper. Since some of the names of the files are the same across the two directories (although the files themselves are not), I thought it would be best if I sent the files attached in two different messages. This message contains the raw and corrected data. 

While I thought the file Index.doc was pretty complete when I wrote it, I see in looking things over that there are a few things that are left unexplained. One is that the file refers to the “problem with the U.S. CPI series.” I honestly don’t recall what that problem was, or how it was corrected. It may just have been some data errors that we (my RAs and I) corrected, or it may have referred to some problem with splicing CPI series when definitions changed. Perhaps you could reconstruct what the problem was, since the files I am sending contain both the corrected and uncorrected U.S. CPI.

The headings in the Excel spreadsheet for CPI data need elaboration. For example, the mnemonics (which are the mnemonics that Datastream uses) for Canada are:

CNOCPCONF - CPI All items

CNOCCPGFF -- CPI All goods excluding food

CNOCCPFDF - CPI Food

CNOCCPSXF -- CPI services excluding rent

CNOCCPRNF - CPI rent

The last four series are mutually exclusive, and collectively comprise the aggregate CPI. (For example, “all goods excluding food” does not mean “all items excluding food.”) In principle, the CPI is a weighted average of these four indexes, though the weights probably vary over time. Since we did not know the actual weights, as we explain in the paper we construct weights (as constants) by regressing the log of the overall CPI on these four items.

I will send the rest of the data as I get a chance. I’m not sure I have documented everything on my computer as well as I documented this CPI data, but hopefully I have.

Section 1, #2

This message contains the rest of the CPI data. It was in the subdirectory “Final” that is referred to in Index.doc, which I sent in the last message.

(Note: For now I do not know how  to get the computer to link to the Gauss matrix files *.fmt in a readable format.)

I am attaching the data used in section II of the paper now. This is the OECD sectoral data. The Appendix describes how we constructed price indexes. The spreadsheets that I am sending show the price index weights that we used. (As the appendix explains, for each country, for each sector we used a constant-weight based on the average weight over our sample.)

For each country there is an Excel file. It is easy to figure out which country the filename refers to.

I’m afraid my documentation of the data for this section is imperfect, but I think I can reconstruct what everything is. First, I have attached a file called ttt_a.xls. This file contains the raw data for all the countries. It also contains the price indexes that were constructed (but see the caveats below concerning Japan and Sweden.) The price data was constructed by dividing series on nominal expenditures by series on real expenditures. The result is included in the spreadsheet. Also, we need a measure of the weight of nontraded goods to do the calculations I do in the JPE paper. Actually, it would be easy to get time-varying weights, because we have all of the data on expenditures. But, to be consistent with the rest of the paper, I assumed constant weights. Those are calculated as the average of the time-varying weights.

For all but UK, the data are: nominal and real expenditure on services; and nominal and real expenditure on commodities. The services data is used to construct the nontraded prices, and the commodities data were used to construct the traded prices.

For the UK, the data was on consumption (nominal and real) of durable commodities, nondurable commodities and services. The services data were used as before. To get the traded price data, we took a weighted average of the durable and nondurable prices. We used constant weights, which again we calculated as the average of the time-varying weights. Those calculations are in the file ukweights.xls. But, the calculation of the UK traded and nontraded price indexes are in the same spreadsheet as the rest of the prices.

In all but this section of the paper I used not seasonally adjusted data. All of this data is seasonally adjusted, except for Sweden and Japan. So to keep things consistent within this section, we seasonally adjusted Sweden and Japan. Here is the rub - I don’t know where the seasonally adjusted data are for these two countries. ttt_a.xls contains price indexes for traded and nontraded for these two countries, but clearly no seasonal adjustment has been done. At some point we seasonally adjusted those series, but I simply cannot find any seasonalized versions of these prices.