RDP 2011-06: Does Equity Mispricing Influence Household and Firm Decisions? Appendix A: Data

Equity prices

I use the Vanguard S&P 500 ETF price measured at the close of trading of each quarter from September 1976 to June 2010. These data are used in all estimation samples and are sourced from Thomson Reuters. For the pre-estimation specification tests that use a sample from March 1952 to June 2010, I backcast (splice) the Vanguard S&P 500 ETF price index using the price changes history for the US S&P 500 (again sourced from Thomson Reuters).

Consumption and labour income

These data are obtained from Martin Lettau's website (available from March 1952 to June 2010 at the time of writing), and are reported in log real per capita terms.^[51] For a full description of these data see Lettau and Ludvigson (2004).

Dividends

Dividends per share are measured as the sum of gross dividends paid in the quarter, with respect to the US S&P 500 index (SPX). Non-seasonally adjusted data are sourced from Bloomberg. Seasonally adjusted estimates are calculated by the author using the US Census Bureau X12 method applied at a quarterly frequency.

Non-equity net wealth

To construct a measure of non-equity net wealth I use:

where:

Wealth data are obtained from the US Flow of Funds Accounts. Total household net wealth is reported under identifier FL152090005.Q, household total equity wealth under FL153064475.Q, all US sector holdings of foreign equity under FL263164103.Q, and all US sector holdings of domestic and foreign equity under FL893064125.Q. In line with Lettau and Ludvigson (2004), all wealth variables are lagged one quarter to be consistent with their beginning of quarter values.

Equity quantities

I use:

where the denominator is the equity prices measure previously described. This measure is also lagged one quarter to be consistent with its beginning of quarter value.

Personal consumption expenditure deflator

Equity prices, non-equity wealth, dividends and labour income are deflated using the personal consumption expenditure deflator, Bureau of Economic Analysis NIPA Table 1.1.9 Line 2.

Population

Non-equity wealth and equity quantities are converted to per capita values using US population estimates, Bureau of Economic Analysis NIPA Table 7.1 Line 18.

Details of the instruments used in estimation are now discussed.

Valuation confidence index

These data are obtained from the Yale School of Management website. I use the proportion of respondents who viewed the stock market as overvalued. From October 1989 to April 2001 biannual survey data are available. From September 2001, six-month-ended averages are reported at a monthly frequency. I construct a biannual measure for the full sample, from October 1989 to April 2010, and then linearly interpolate the data to a quarterly frequency.^[52] As noted in the main text, the use of a linear interpolation will have no effect on the validity of this instrument once the second difference of this measure is used in estimation.

Option volatility

This measure is 30-day option-implied equity volatility with respect to the US S&P 100. These data are sourced from Bloomberg (with ticker VXO) and are measured at market close on the last trading day of the relevant quarter. For comparison, estimation on a shorter sample was also undertaken using 30-day option-implied equity volatility with respect to the US S&P 500. These data are also sourced from Bloomberg (with ticker VIX).

Forecast dispersion

This measures the weighted standard deviation in long-term earnings-per-share-growth forecasts for companies in the US S&P 500, using the Institutional Brokers' Estimate System (I/B/E/S). The weights used reflect market capitalisation, with this series sourced from Thomson Reuters. This series is measured at the beginning of the quarter.

Footnotes

See http://faculty.haas.berkeley.edu/lettau/data/cay_q_10Q2.txt. [51]

From September 2001 onwards, the Yale School of Management reports the six-month-ended average percentage responses, . To adjust for this change in reporting, biannual monthly survey responses are calculated by the author using

View MathML

[52]