House prices, equity prices, investment and credit--what do they tell us about banking crises? A historical analysis based on Norwegian data.

Author:Riiser, Magdalena D.
 
FREE EXCERPT

In recent years, many countries have experienced a sharp rise in house prices and household credit. Many have expressed concern that this development is not sustainable over time and that it may lead to financial imbalances. In this article, we will consider whether historical indicators can predict banking crises through the last 150 years. Using a Hodrick-Prescott filter, we calculate the gap between actual observations and trend for real house prices, real equity prices, gross fixed investment and credit on the basis of Norwegian data back to 1819. We find that all gap indicators are useful in predicting earlier banking crises in Norway. With few exceptions, the indicators show a common pattern - the gaps widen from one to six years prior to the banking crises and subsequently fall. As a rule, at least two of the gap indicators have high values prior to the banking crises, indicating that combinations of indicators may increase the strength of the analysis. We also find that indicator values that can be associated with a banking crisis, i.e. the threshold values, may be somewhat higher in Norway than in comparable international studies.

1 Introduction

In recent years, many countries have experienced strong increases in house prices and household credit. Many have expressed concern that this development is not sustainable over time and that the "borrowing bubble" may burst. A number of studies have presented economic indicators that can predict banking crises. In this article, we look at some of these indicators for Norway. Using data that go back to 1819, we try to reveal whether there are recurring relationships between some economic variables and banking crises in Norway.

This article is organised as follows: Section 2 discusses the relationship between credit, asset prices and banking crises and provides a brief summary of international studies. Section 3 presents different indicators for Norway and considers the relationship between these indicators and banking crises as far back as the 1800s. Section 4 summarises our findings.

2 The relationship between credit, asset prices and financial stability

One hypothesis about the causes of banking crises is the hypothesis of financial fragility, which is investigated in a number of studies, including those conducted by Minsky (1977) and Kindleberger (1978, 2000). According to this hypothesis, considerable optimism in periods of economic expansion can push up both asset prices and investment and result in high credit growth. This may contribute to the build-up of financial imbalances. In the event of disturbances in the economy, optimism will wane. Asset prices and investment will fall. The quality of banks' portfolios will be put to the test and the value of banks' collateral will diminish. Servicing debt will become a problem and banks' loan losses will increase.

Recent studies focus on equity prices as an indicator of impending banking crises. These studies show that equity prices rise sharply and then fall for up to a year before a banking crisis. (1)

A large portion of the literature is devoted to the importance of credit for banking crises. (2) The main conclusion is that strong growth in domestic credit increases the probability of financial instability. Most studies concerning credit place emphasis on growth during a limited time period. For example, they consider the implications of high credit growth for a period of one year. Consequently, stock variables and cumulative processes are virtually disregarded. Meanwhile, the vulnerability of the non-financial sector (non-financial enterprises, households and municipalities) will not only depend on debt growth, but also on the level of debt. Strong credit growth for a period of some years, from an initially low level, will not necessarily represent a threat to debt-servicing capacity.

History shows that a number of factors and events have usually played a part in triggering financial instability. The studies generally reveal relationships between developments in asset prices and credit on the one hand and financial distress on the other. However, they provide few numerical indicators which may be used by central banks and government authorities to assess whether or not financial stability is at risk.

Borio and Lowe (2002) discuss these problems. In their study, they look at real asset prices, credit to the private sector and investment. They focus on cumulative processes. To capture such effects, they analyse developments in credit and investment as a percentage of GDP instead of looking at growth rates over a shorter time period. The indicator for credit as a percentage of GDP is hereafter referred to as the credit gap. This is compared with an indicator for growth in inflation-adjusted credit in order to examine the predictive powers of indicators linked to level compared with pure growth indicators.

The primary objective is to construct indicators that can predict banking crises. The idea, which is based on Kaminsky and Reinhart (1999), is to find a threshold value for each of the indicators which can signal financial problems. The method involves calculating a gap for the variables concerned, defined as the deviation between actual observations and a trend. The gaps are calculated as a per cent of the trend with the exception of the credit gap, which is measured in percentage points.

Borio and Lowe (2002) examine both single indicators and combinations of indicators. They also look at multiple horizons and consider the usefulness of indicators in predicting banking crises within one, two and three years. They use data from 34 countries with a total of 38 banking crises during the period 1960-1999.

Of the four indicators examined, the credit gap provides the best results. A gap of 4 percentage points predicts nearly 80 per cent of the banking crises within one year and gives false signals in only 18 per cent of the cases. The credit gap is clearly a better indicator than the gap in credit growth. The predictive powers of the gaps in real equity prices and investment as a percentage of GDP are lower than that of the credit gap. In addition, these two gap indicators are fairly noisy. Another finding from the study is that expanding the time horizon improves the predictive powers of the indicators, in particular the indicators for real equity prices and credit.

Borio and Lowe (2002) experiment with various combinations of indicators and find that this improves the predictive properties. They conclude that the combination of a credit gap with a threshold value of 4 percentage points and a real equity price gap with a threshold value of 40 per cent provides the best results. Including the investment gap does not increase the predictive powers of the indicators. Expanding the time horizon from one to three years improves the indicators' predictive powers.

In Borio and Lowe (2004), the analysis is expanded by using quarterly data and extending the time horizon to three-to-five years. The predictive powers of the indicators improve compared with the authors' previous study.

3 House prices, equity prices, investment and credit in Norway

3.1 Calculating gap indicators for Norway

We have used the method described in Borio and Lowe (2002) to test the hypothesis of financial fragility on historical data for Norway. We have calculated the gap in real house prices, real equity prices, investment as a percentage of GDP and credit as a percentage of GDE The gaps are measured as percentage deviations from the trend, with the exception of the credit gap, which is measured as a percentage of GDP, and here we use the difference in percentage points from the trend. We, like Borio and Lowe (2002), calculate the trend using a Hodrick-Prescott filter (HP filter) (3) and a recursive method. (4) This means that only data up to the beginning of each year is included in the calculation of the trend value for this year. This implies that we analyse the same information that was in principle available to decision-makers at any given time. (5)

We use data from as far back as 1819 from Norges Bank's historical monetary statistics. (6) We include an indicator for house prices. (7) House prices have rarely been used in similar studies because it has been difficult to find adequately long time series for property prices (house prices and prices for commercial property) which are comparable across countries. The close relationship between house prices and household credit (8) and the importance of house prices for banks' collateral make it very interesting to include them in the analysis. Our method for finding the indicators' threshold values differs somewhat from the method used by Borio and Lowe (2002). Since our study involves only one country, we use the peaks in the gaps to establish the threshold values, whereas Borio and Lowe have panel data and weigh the number of predicted crises...

To continue reading

REQUEST YOUR TRIAL