Estimating liquidity premia in the Spanish government securities market

Abstract

This paper investigates the presence of liquidity premia in the relative pricing of assets traded on the Spanish government securities market. First, a classification of bonds into four different categories based on their degree of liquidity is proposed. Second, liquidity premia are estimated introducing liquidity parameters in the estimation of the zero-coupon yield curve. Results suggest the existence of a liquidity premium for post-benchmark bonds (both strippable and non-strippable). The size of this premium is relatively small. In the case of pre-benchmark bonds, the lack of liquidity does not seem to be priced. It is also shown that these pricing discrepancies are robust to the impact of taxes on bonds.

Keywords:

• liquidity premium
• bid-ask spread
• yield curve
• benchmark

Acknowledgements

The authors wish to thank Juan Ayuso, Clemente Fernández, Víctor García-Vaquero, José Ramón Martínez, Alfredo Martínez, Jorge Martínez, Soledad Núñez, Oreste Tristani, Christian Upper and participants in the BIS' Workshop on Market Liquidity and in the seminar at Banco de España for helpful comments and suggestions. Valuable comments from two anonymous referees are also gratefully acknowledged.

Notes

Note that the bid–ask only captures the tightness, which is one of three dimensions of liquidity. The two other are depth and resiliency. See Kyle (1985), Black (1971) or BIS (1999) for further discussion.

See, for example Amihud and Mendelson (1986, 1989).

Strips are not included. The percentage of stripped bonds is very low in the Spanish market.

This information is provided by the Central de Anotaciones of the Banco de España.

Four in the case of short-term instruments.

In the information provided by Reuters it is not uncommon to find negative bid–ask spreads, possibly due to the lack of contemporaneity in quotations.

In fact, in the case of the turnover ratio, no significant differences are observed between pre-benchmark and benchmark issues.

For this analysis we use the Reuters-BDE database because it provides a spread measure more closely linked to the concept of market spread (see Section 3 for a description of this database and a comparison with the Bloomberg data).

BIS (1999) compares bid–ask spreads for ‘on-the-run’ and ‘off-the-run’ issues for Canada, Italy, Japan, UK and US. In all cases they find the same results as ours, except for Japan, where the liquidity of on-the-run 10-year bonds makes their spread narrower than the one for ‘on-the-run’ 5-year bonds. (See Table 3 of the document Market Liquidity: Research Findings and Selected Policy Implications).

They use non-linear least squares to fit a cubic spline to the after-tax cash flows of bonds.

See Svensson (1994).

Henceforth, all the estimation results are obtained using the Bloomberg dataset. When we use the Reuters-BDE dataset we obtain similar results.

Elton and Green (1998) use instead a quantitative variable (volume) to proxy for liquidity.

We use the mid-point between the quoted bid and ask prices. For short-term we use repo and bill rates.

This model is a particular case of the Svensson model, where the instantaneous forward rate is modelled with just one hump, i.e. β₃ and τ₂ are set to zero and one respectively.

We concentrate on the interpretation of parameters γ₀, γ₁, γ₂, which will be reported in different tables. The remaining parameters of the Svensson and Nelson–Siegel model are presented in Table 5 for all the models estimated in the paper.

An anonymous referee raised the issue of whether similar reasoning for benchmark bonds could lead to smaller relative premia between benchmark and non-benchmark issues when the former were to acquire post-benchmark status in the near future. This does in fact seem to hold around the period in which the benchmark status of the issues changes. In these periods, which last around two months, the liquidity premium of non-benchmark bonds relative to benchmark bonds is generally not statistically different from zero (see shaded areas in Fig. 3).

These transactions consist of selling non-strippable bonds to non-resident investors, who are not subject to withholding tax, before the date of the coupon, and buying them back again after that date.

From 4/01/99 to 30/9/99 we estimate the Nelson–Siegel model, and from 1/10/99 to 14/4/00 the Svensson model.

This information is available on request.

The only exception being those bonds that mature within the sample period. The low duration of these bonds implies that small differences in quoted and traded prices translate into significant discrepancies in terms of yield to maturity.