اثر سیاست های پولی در رشد قیمت واقعی مسکن در جنوب آفریقا: یک عامل تقویتی بردار رگرسیون خودکار (FAVAR)

This paper assesses the impact of monetary policy on real house price growth in South Africa using a factor-augmented vector autoregression (FAVAR), estimated using a large data set comprising of 246 quarterly series over the period 1980:01 to 2006:04. The results based on the impulse response functions indicate that, in general, house price inflation responds negatively to monetary policy shock, but the responses are heterogeneous across the middle-, luxury- and affordable-segments of the housing market. The luxury-, large-middle- and medium-middle-segments are found to respond much more than the small-middle- and the affordable-segments of the housing market. More importantly, we find no evidence of the home price puzzle, observed previously by other studies that analyzed house prices using small-scale models. We put this down to the benefit gained from using a large information set.

The recent global economic downturn attributed to the sub-prime crisis in the US with rapid contagion worldwide, particularly in the housing sector, has attracted the attention of academics, policymakers, and economic agents at large. Stock and Watson (2003) pointed out that housing prices are leading indicators for real activity, inflation, or both, and, hence, can serve as an indicator as to where the real economy is heading. Evidence in the recent literature, for example, Iacoviello, 2005, Case et al., 2005, Iacoviello and Neri, 2008, Vargas-Silva, 2008a and Vargas-Silva, 2008b amongst others, show a strong link between the housing market and economic activity in the US. Moreover, the recent emergence of boom–bust cycles in house prices have been an issue of concern for policy markers (Borio et al., 1994, Bernanke and Gertler, 1995 and Bernanke and Gertler, 1999), since the bust of the house price bubble is always followed by significant contractions in the real economy (Iacoviello and Neri, 2008). Given this, it is crucial for central banks to analyze thoroughly the effects of monetary policy on asset prices in general, and real estate in particular, which, in turn, would lead to the understanding of effects of policy on the economy at large. In this backdrop, this paper assesses the impact of monetary policy shocks on real house price growth, i.e., the growth rate of the ratio of nominal house price to the Consumer Price Index (CPI), for the luxury, large-, medium- and small-middle-segments and affordable housing for the South African economy1 by exploiting a data-rich environment that includes 246 quarterly series, such as income, interest rates, construction costs, labour market variables, stock prices, industrial production, and consumer confidence index over the period 1980:01 to 2006:04. For this purpose, the framework used in this paper is a factor-augmented vector autoregression (FAVAR) model proposed by Bernanke et al. (2005). As Bernanke et al. (2005) indicate, monetary authorities analyze literally thousands of variables in their decision-making process, hence, it is aberrant for anyone, who tries to mimic actions of a central bank, to ignore this fact. Furthermore, the recent literature (Stock and Watson, 2004, Rapach and Strauss, 2007, Rapach and Strauss, 2009, Das et al., 2008, Das et al., 2009 and Das et al., forthcoming) provide evidence of the fact that numerous economic variables are potential predictors of house price growth. Intuitively, the FAVAR approach boils down to extracting a few latent common factors from a large matrix of many economic variables, with the former maintaining the same information contained in the original data set without running into the risk of the degrees of freedom problem.2 Note, the motivation to use the three major segments of the housing market, with the middle-segment subdivided further into three categories based on sizes, and not just the aggregate housing market, emanates from the fact that the market for different house-types are found to behave differently (Burger and van Rensburg, 2008). Clearly then, the impact of monetary policy on the different segments of the South African housing market is less likely to be homogenous. This is more so, when one realizes that different housing segments cater to different income-groups. To the best of our knowledge, this is the first study to analyze the effect of monetary policy on real house price growth in South Africa using a FAVAR. The only other paper that deals with the impact of monetary policy on the South African housing market is that by Kasai and Gupta (2008). The authors investigated the effectiveness of monetary policy on house prices in South Africa, before and after financial liberalization, with financial liberalization being identified with the recommendations of the De Kock Commission (1985). Using both impulse response and variance decomposition analysis performed on three-variable structural VARs (SVARs), comprising of the growth rate of the real GDP, house price inflation and the Treasury Bill rate, estimated separately on the three categories of the middle-segment of the housing market, the authors found that irrespective of house sizes, during the period of financial liberalization, interest rate shocks have had relatively stronger effects on house price inflation. But, given that the size of these effects were nearly negligible, the result seems to indicate that house prices are exogenous, and, at least, are not driven by monetary policy shocks. Though insightful, the paper by Kasai and Gupta (2008), just like Iacoviello, 2002, McCarthy and Peach, 2002, Iacoviello and Minetti, 2003 and Iacoviello and Minetti, 2008,3Vargas-Silva (2008a), is based on a small-scale model, which, in turn, limits it to only three variables. In fact, all the other studies, being based on either reduced-form Vector Autoregressive (VAR), Vector Error Correction (VEC), SVAR or DSGE models, could handle at most 8 to 12 variables only. Arguably, and as indicated above, there are a large number of variables that affects monetary policy and the housing market, and not including them often leads to puzzling results, for example the homeprice puzzle 4 in McCarthy and Peach, 2002 and Kasai and Gupta, 2008, 5 which are not in line with economic theory due to the small information set ( Sims, 1992 and Walsh, 2000). Moreover, in these studies, the authors often arbitrarily accept specific variables as the counterparts of the theoretical constructs (for example the gross domestic product as a measure of economic activity or the first difference of the logarithm transformed consumer price index as a measure of inflation), which, in turn, may not be perfectly represented by the selected variables. In addition, previous studies can only obtain the impulse response functions (IRFs) from those few variables included in the model, implying that in each VAR, VECM, SVAR or DSGE, the IRFs are typically obtained with respect to only one variable related to the housing market. Given its econometric construct, the FAVAR model addresses all these problems. The remainder of the paper is organized as follows: Section 2 briefly discusses the FAVAR framework, while, Section 3 discusses the data and the identification structure. Section 4 reports and analyzes the impulse response functions, and Section 5 concludes.

This paper assesses the impact of a positive monetary policy shock on real house price growth for the five segments of the South African economy using a FAVAR estimated with 246 variables spanning the period of 1980:Q1 to 2006:Q4. Overall, the results show that real house price growth responds negatively to a positive monetary policy shock, suggesting that the framework does not experience the home price puzzle, encountered by Kasai and Gupta (2008) while analyzing monetary policy shocks with three-variable SVARs for the middle-segment of the South African housing market. This result points to the benefit gained by using a large information set. Not surprisingly, the reaction of real house price growth rate is found to differ across the five housing categories, indicating the segmented nature of the market. Specifically, we find the luxury, the large-middle- and the medium-middle-segments to experience the biggest negative impacts following a contractionary monetary policy. However, unlike the two middle-segments, the effect on luxury housing recovers back to its original level much faster. The lower end of the market, i.e., the small-middle-segment and affordable housing witness small and short-lived negative effects. As part of future research, it would be interesting to analyze the robustness of the results based on a large-scale Bayesian VAR (LBVAR), developed recently by Banbura et al. (2008), since just like the FAVAR, the LBVAR, given its estimation methodology, can also handle a data set of any size. Moreover, unlike the FAVAR, the LBVAR, via appropriate design of the interaction matrix of the variables, can account for spatial influences of neighboring regions and also asymmetric effects of regional variables and national variables on each other. Note, regional variables are likely to have minor effects on national variables, while, the national variables are more prone to affect the regional variables strongly. However, given that regional (provincial) level house price data in South Africa is only limited to the middle-segment, we would have to restrict our analysis to this section of the housing market only. Nevertheless, a regional analysis would be worth the endeavor in understanding which province(s) in South Africa plays an important role in determining the dynamics of the national house price. Finally, given that the Bayesian methodology does not require us to ensure stationarity of variables, we can analyze house prices at levels rather than their growth rates, if necessary.