Quarterly Bulletin
Economic Modelling Articles

Addendum to Report on modelling and forecasting at the Bank of England (166k)
The Spring 2003 Quarterly Bulletin contained a report on modelling and forecasting at the Bank of England by Adrian Pagan. This article is a postscript to Professor Pagan’s original report, and covers the introduction of the Bank’s new macroeconomic model (the Bank of England Quarterly Model, or BEQM). The overall assessment is broadly positive, and the Bank would again like to thank Professor Pagan for the valuable insights that his report has provided.

Credit channel effects in the monetary transmission mechanism (97k)
(By Simon Hall of the Bank's International Finance Division). Economic models often assume for simplicity that the impact on the wider economy of changes in financial conditions can be summarised by a relatively limited set of financial variables, such as short-term risk-free interest rates and long-term government bond rates. However, financial developments can, at times, have important effects on the economy, which these variables would not necessarily indicate. For example, following the suspension of debt payments by Russia in the summer of 1998 and the emergence shortly afterwards of problems at the hedge fund Long Term Capital Management (LTCM), interest rates on corporate debt rose relative to rates on government debt, and a number of central banks reduced official interest rates to mitigate possible effects on spending in the wider economy. In practice, policy-makers take account of a wide range of information on conditions in financial markets to monitor, and potentially respond to, these sorts of developments.

This article reviews potential theoretical explanations for two features of finance provision—the apparent preference by many borrowers to finance spending using own funds, and for many of those who do borrow, to rely on bank rather than capital market finance. These so-called 'credit channel' models help to explain why borrowers' financial positions might affect their spending, and why shocks to banks can have a marked impact on borrowers that are particularly dependent on bank finance. As such, these models illustrate some important interactions between the monetary and financial stability objectives of central banks and highlight the need for policy-makers to monitor a wide range of financial indicators.

In practice, banking system distress and significant disruptions to bank loan supply are relatively rare in developed banking sectors, as in the United Kingdom. As such, bank lending credit channel effects may be relatively infrequent. Balance sheet credit channel effects probably play a more continuous role in the economy, but they too will likely vary in strength over time, reflecting structural changes in the financial system and cyclical fluctuations in borrower financial health. This article focuses on a representative model of balance sheet effects. Two other articles in this Bulletin use the framework of this model to show how credit channel effects may affect spending in the UK corporate and household sectors.

Economic models at the Bank of England (96k)
In April 1999, the Bank of England published Economic models at the Bank of England, a book that described the economic modelling tools that help the Monetary Policy Committee (MPC) in its work. It was made clear at the time that economic models should not be thought of as fixed in form or content, and that model development is a continual process. An update to the book was published in September 2000 covering model developments over the past 18 months, particularly in relation to the Bank's main macroeconometric modelling tool. The update, while giving details of the core macroeconometric model, also refers to other work within the Bank that has added to the range of models used by the MPC.

New estimates of the UK real and nominal yield curves (83k)
(by Nicola Anderson and John Sleath of the Bank's Monetary Instruments and Markets Division). This article presents some new improved estimates of the UK yield curve, both nominal and real. It describes the rationale for changing the estimation techniques that we have previously used, in the light of our own experience and developments in the academic literature. The article also illustrates the use of data from the general collateral repo market to derive estimates of the nominal yield curve at short maturities.

Nominal yield curves have been estimated in the Bank for more than 30 years. For the past five years, in common with many other central banks, we have used the estimation method proposed by Svensson (1994, 1995). This is a parametric method, with the entire curve described by a single set of parameters representing the long-run level of interest rates, the slope of the curve and humps in the curve. Previously we used an in-house non-parametric method described by Mastronikola (1991). And before that we used another parametric approach, with the parameters reflecting, among other things, segmentation in the market and the planning horizons of different investors.

Estimation of the real yield curve is a more recent innovation, made possible by the introduction of index-linked bonds in the United Kingdom in 1981. As these bonds are indexed only imperfectly to the price level, we have to use information from the nominal yield curve to extract the real risk-free rates of interest embodied in their prices. Until now we have been using an iterative technique developed by Deacon and Derry (1994), in which the real yield curve is described by a restricted version of Svensson's model.

As discussed by Breedon (1995), the Svensson method was preferred both to the earlier in-house method and the range of alternative options available at the time, on the basis of three key criteria. Specifically:

• the technique should aim to fit implied forward rates (rather than, for example, yields), since the final objective is to derive implied forward rates;
• it should give relatively smooth forward curves, rather than trying to fit every data point, since the aim is to supply a market expectation for monetary policy purposes, rather than a precise pricing of all bonds in the market; and
• it should allow as many economic restrictions as possible to be imposed.

For maturities of less than two years, estimates of both the real and nominal yield curves have not been thought to be reliable, and as a result have not been used by the Bank's Monetary Policy Committee, nor published in the Inflation Report or Quarterly Bulletin. This is partly because there are few gilts at the short end of the yield curve (ie with terms to maturity of two years or less), where expectations may be relatively precise and where the curve may be expected to have quite a lot of curvature. More recently, experience has led us to question whether the Svensson estimates, even at the longer maturities, are the best guide to monetary conditions in the United Kingdom.

The opportunity to shed new light on the performance of these models has arisen, partly through the relatively recent arrival of additional information from the gilt market (in the form of strips prices), and partly through the development of new techniques for estimating the yield curve. In the latter case, we find that a new model developed by Waggoner (1997) offers a number of improvements on the parametric methods currently used to estimate both the real and nominal yield curves. In addition, improvements in extracting the real yield curve from index-linked bond prices can be found using the non-iterative technique developed by Evans (1998).