2 Dataset and Empirical Model
The dataset consists of US monthly aggregate data ranging from March 1973
to June 2006. The data on issuance of bonds and stocks are taken from the Statisti-
cal Supplement to the Federal Reserve Bulletin. Figures represent gross proceeds
of issues maturing in more than one year, and include both financial and nonfi-
nancial corporations, and both private and public placements in the case of bonds,
while public placements only in the case of shares.3 All figures are deflated by
using the Consumption Price Index. The dataset includes also the series of the US
industrial production index, while the stock and bond market indices considered
are the Standard&Poor500 (S&P500) and the Lehman Brothers Corporate Bond
Index (LBCB).4
The basic relations we want to test involve linear relationships among the volumes
raised by means of primary placements of shares (St) and corporate bonds (Bt),
plus a set of predeterminated variables which includes stock (RS,t) and bond (RB,t)
market returns, their volatility (σRS,t and σRB,t ), and the growth rate of the US in-
dustrial production index (yt). Stock and bond market returns are computed by
employing monthly S&P500 and LBCB. Stock and bond market volatilities are
modelled by fitting ARMA-GARCH processes to the same series.5
3Figures exclude secondary offerings, employee stock plans, investment companies other than
closed-end, intra-corporate transactions, and Yankee bonds.
4These series are obtained from the FRED database at the Federal Reserve Bank of St Louis
and Datastream.
5The best fitting models for monthly stock and bond market returns are respectively a
ARMA(1,1)-GARCH(1,1) and a ARMA(1,0)-GARCH(1,1). These specifications deliver stan-
dardized residuals and squared residuals not serially correlated. Moreover, ARCH LM tests sug-
gest the absence of GARCH effects in standardized residuals. To save space these results are not