Investment choice and asset allocation of Italian households: the discrete-continuous approach

Abstract

Financial and monetary policies are considered to be effective depending on the reaction of financial markets which are increasingly populated by households. In this article, from intertemporal settings, I derive a Financial Almost Ideal (FAI) Demand System and I estimate it by highlighting the determinants of both limited participation to financial markets and asset substitutability/complementarity in the allocating stage. Finally, the wealth elasticities provide some further insights on the low diffusion of the newer and more complex financial instruments across Italian households.

Acknowledgements

I wish to express my deepest gratitude to Carlo Andrea Bollino and Luca Pieroni for the useful comments. I am also grateful to Michele Bagella, Gustavo Piga and all the participants to my seminar held at University of Tor Vergata–24 January 2006. Finally, I would like to thank Guglielmo Weber, Giuseppe De Luca and all the participants to the Second Italian Congress of Econometrics and Empirical Economics (Rimini, 25–26 January 2007) for the helpful discussions. All the remaining errors are my own responsibility.

Notes

¹ Several arguments may be called for justifying the presence of asset shares into utility function. Most of these arguments refers to the so-called motives approach which regards financial assets as providers of transactional, precautionary or speculative services and, in turn, as yielding utility to their holders.

² These technical requirements are necessary to ensure that the value function can be written with a recursive form.

³ For instance, in Barr and Cuthberson (1991) and Dinenis and Scott (1993), the financial asset demand system is specified in terms of expected log-asset prices, i.e. . Hence, in a parameterization working with log-compounded returns, one may obtain economic interpretations in log-expected price terms by inverting the sign of the estimated coefficients.

⁴ Several examples may be provided to clarify this statement. For instance, the bequest motive can affect not only the saved amount, but also the composition of the household portfolio and the same can be claimed for the precautionary and life-cycle motives. However, the most striking example is clearly offered by the improvement motive: the gains from the invested wealth reflect the underlying portfolio allocation rather than the decisions on the amount to save.

⁵ In particular, the household investment forms are articulated in: (1) Bank Deposits, Certificates of Deposits, Repos; (2) Post Office Deposits; (3) Italian Government Securities; (4) Bonds, Shares of Italian Mutual Funds; (5) Italian Shares; (6) Managed Savings; (7) Foreign Securities (issued by nonresidents) and (8) Loans to Cooperatives.

⁶ The first round is a subsection of the survey that is exclusively addressed to note down peculiar aspects of households’ economics and that involves only a random subsample of the households.

⁷ The re-classification that I propose neglects the loans to cooperatives which represent an insufficiently widespread investment alternative and other bank and postal products, including certificates of deposits, repos and PO savings certificates which are held sporadically in the household portfolios.

⁸ In the survey, only the rate earned on bank deposits is reported. This is assumed to be the rate of the aggregate bank and postal deposits.

⁹ In particular, theoretical gain is denned as the gain obtained from selling the asset at the end of 2002 (see questions C46, C47, C48, C49).

¹⁰ Even if such a way to quantify asset returns may penalize those assets which provide the investors with systematic stream of payments, it represents, however, an immediate measuring of financial gains/losses arising from the survey. Alternative ways to avoid such assumption is to find out a reasonable relationship that can be used to charge the bulk of interests, coupons and dividends to the corresponding asset returns.

¹¹ Dependence among participation decisions are neglected in the two-step estimator of Yen et al. (2002), where Σ_ηη = I ₄ entails that system of participation equations can be estimated by four different univariate probits.

¹² For instance, in multinomial probit with J alternatives, the number of the parameters contained in the unrestricted variance–covariance matrix is given by: J(J + 1)/2, while the maximum number of identified variance–covariance parameters is given by: J(J − 1)/2 − 1.

¹³ For an exhaustive survey on this topic, see Bunch and Kitamura (1991).

¹⁴ Jointly testing the adding-up and symmetry yields: χ²(9) = 6.68 (p-value: 0.67).