The Spillover Effects of a Downturn in China’s Real Estate Investment
Author: Mr. Ashvin Ahuja and Ms. Alla Myrvoda1
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

Author’s E-Mail Addresses: aahuja@imf.org
Publication Date:
05 Nov 2012
eISBN:
9781475549003
Language:
English
Keywords:
WP; China; investment; price; China investment bust; price inflation; investment bust; sovereign bond; near-term price expectation; capital goods
Download PDF (4.1 MB)

Real estate investment accounts for a quarter of total fixed asset investment (FAI) in China. The real estate sector’s extensive industrial and financial linkages make it a special type of economic activity, especially where the credit creation process relies primarily on collateral, like in China. As a result, the impact on economic activity of a collapse in real estate investment in China—though a low-probability event—would be sizable, with large spillovers to a number of China’s trading partners. Using a two-region factor-augmented vector autoregression model that allows for interaction between China and the rest of the G20 economies, we find that a 1-percent decline in China’s real estate investment would shave about 0.1 percent off China’s real GDP within the first year, with negative spillover impacts to China’s G20 trading partners that would cause global output to decline by roughly 0.05 percent from baseline. Japan, Korea, and Germany would be among the hardest hit. In that event, commodity prices, especially metal prices, could fall by as much as 0.8–2.2 percent below baseline one year after the shock.

Abstract

Real estate investment accounts for a quarter of total fixed asset investment (FAI) in China. The real estate sector’s extensive industrial and financial linkages make it a special type of economic activity, especially where the credit creation process relies primarily on collateral, like in China. As a result, the impact on economic activity of a collapse in real estate investment in China—though a low-probability event—would be sizable, with large spillovers to a number of China’s trading partners. Using a two-region factor-augmented vector autoregression model that allows for interaction between China and the rest of the G20 economies, we find that a 1-percent decline in China’s real estate investment would shave about 0.1 percent off China’s real GDP within the first year, with negative spillover impacts to China’s G20 trading partners that would cause global output to decline by roughly 0.05 percent from baseline. Japan, Korea, and Germany would be among the hardest hit. In that event, commodity prices, especially metal prices, could fall by as much as 0.8–2.2 percent below baseline one year after the shock.

I. Introduction

Real estate investment accounts for a quarter of total fixed asset investment (FAI) in China. The real estate sector’s extensive industrial and financial linkages make it a special type of economic activity, especially where the credit creation process relies primarily on collateral, like in China. As a result, the impact on economic activity of a collapse in real estate investment in China—though a low-probability event—would be sizable, with large spillovers to a number of China’s trading partners. Using a two-region factor-augmented vector autoregression model that allows for interaction between China and the rest of the G20 economies, we find that a 1-percent decline in China’s real estate investment would shave about 0.1 percent off China’s real GDP within the first year, with negative spillover impacts to China’s G20 trading partners that would cause global output to decline by roughly 0.05 percent from baseline. Japan, Korea, and Germany would be among the hardest hit. In that event, commodity prices, especially metal prices, could fall by as much as 0.8–2.2 percent below baseline one year after the shock.

The relatively new private property market in China has always been susceptible to excessive price growth, requiring escalated intervention by the authorities over the years. The underlying structural features of the economy, namely low real interest rates in a high growth environment, the under-developed financial system (offering few alternative assets) and a closed capital account, foster overinvestment in real estate and create an inherent tendency for bubbles in the property market, posing risks to market sustainability and financial stability. Currently, real estate investment accounts for one quarter of China’s fixed asset investment. It has been growing at around 30 percent per annum over the past two years (2010–2011)

A01ufig01

Fixed Asset Investment: by Industry

(In percent of total, 2011)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

Source: CEIC
A01ufig02

Property Price and Real Estate Investment

(In percent, year-on-year growth)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

Policy response relies largely on quantity-based tools, the effectiveness of which tends to erode over time as more transactions are intermediated outside of the banking system, requiring more potent policy responses. In the most recent episode of property boom, which started around mid-2009, the authorities escalated its response with restrictions of second and third home purchases in larger cities and credit limits on property developers. Thus far, the authorities appear to have succeeded in curbing market exuberance while maintaining robust investment growth, chiefly through an expansion of social housing programs and a selective easing of financial conditions for first-time home buyers. Nevertheless, developers’ financial conditions are deteriorating, and there is a tail risk that policy over-tightening could turn near-term price expectation decidedly negative as high inventory-to-sale ratios compress developers’ profitability further, leading to a collapse in real estate investment.

A01ufig03

Property Prices

(In percent, year-on-year growth)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

Source: CEIC; Soufun; and IMF staff calculations.

The risk to growth and financial stability of a collapse in real estate investment is high, based on the expected economic repercussion should that event come to pass. The analysis based on China’s input-output data shows that the real-estate-dependent construction industry, which accounts for 7 percent of GDP, creates significant final demand in other domestic sectors; that is, it has among the highest degrees of backward linkages, particularly to mining, manufacturing of construction material, metal and mineral products, machinery and equipment, consumer goods, as well as real estate services. Moreover, real estate is used principally as collateral for external financing of private and state-owned enterprises as well as local government’s investment projects, and other economic activities. As a result, a decline in real estate investment has the potential to disrupt the production chain throughout China’s economy, and with that a potential for external spillover to G20 trading partners.

A01ufig04

Backward Linkages: Selected Contributors

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

II. Modeling the Spillover Effects

We use a factor-augmented VAR (FAVAR) approach pioneered by Bernanke, Boivin and Eliasz (2005) to gauge the domestic and global spillovers of a slowdown in China’s real estate investment in an event of a sharp property market correction. Following Boivin and Giannoni (2008), the FAVAR framework is extended into a two-region model that allows China to interact with the rest of the world (represented in this experiment by the other G20 economies). The analysis captures the feedback from China to the rest of the world, and vice versa, over time. It also captures the spillover effect between the rest of the G20 economies from a specific event originated in China.

The fact that market participants monitor hundreds of economic variables in their decision making process provides motivation for conditioning the analysis of their decisions on a rich information set. The FAVAR framework extracts information from the rich data set to gauge the impact of particular forces that may not be directly observable. These “forces” are treated as latent common components, which are inter-related, and their impacts on economic variables are traced through impulse response functions. By accounting for unobserved variables, there is a better chance that findings based on spurious association can be avoided.

More detailed description of the model and estimation strategy can be found in the appendix. Briefly, the model is a stable FAVAR in growth (except for balances and interest rates) with 5 common factors for each region (China and the rest of the G20 economies) and China’s real estate investment. The model uses one lag. The Cholesky factor from the residual covariance matrix is used to orthogonalize the impulses, which imposes an ordering of the variables in the VAR and treats real estate investment as exogenous in the period of shock. The results are robust to re-ordering within factor groups. The data set is a balanced panel of 390 monthly time series from the G20 stretching from 2000M1 to 2011M9, with 68 China’s variables and 322 from the rest of the world (see data description, transformation, and sources in Appendix B). Our sample contains at least one full cycle of real estate investment and property market in China. It covers the period when China entered the WTO and became increasingly integrated with the world economy.

Since the model is in growth, the experiment assumes an exogenous, temporary, one-standard-deviation growth shock to China’s real estate investment. The shock dampens within a few months and dissipates fully after around 36 months. Specifically, this is a onetime 49-percentage-point (seasonally adjusted, annualized) drop in real estate investment growth that reverts to trend growth largely within 4-5 months.2 While this is a temporary, negative growth shock, the decline in real estate investment level is permanent. The shock is approximately equivalent to a 2-percent drop from baseline in real estate investment level 12 months after. The analysis does not assume policy response beyond that which was already in the sample.

article image

Twenty-four-month peak impacts to one-standard-deviation shock to real estate investment are reported with standard error bands in the charts below. Impacts on levels 12 months after the shock, in percent below baseline, are also derived and reported for comparison in Tables 1-4.

Table 1.

Impacts one year after a 1-percent exogenous decline in China’s real estate investment: Selected China Indicators

article image
Remark: A one-standard-deviation decline in growth is equivalent to 2-percent decline in real estate investment levels from baseline
Table 2.

Impacts one year after a 1-percent exogenous decline in China’s real estate investment: Economic Activity Indicators

(In percent, year-on-year)

article image
Remark: A one-standard-deviation decline in growth is equivalent to 2-percent decline in real estate investment levels from baseline.

Estimate for Australia is not statistically significant.

Canada’s economic activity is represented by monthly real GDP Index, all industries.

China’s industrial sector activity is represented by gross industrial value added.

Table 3.

Impacts one year after a 1-percent exogenous decline in China’s real estate investment: Trade Indicators

(In percent, year-on-year)

article image
Remark: A one-standard-deviation decline in growth is equivalent to 2-percent decline in real estate investment levels from baseline.*Import-weighted. ** Export-weighted.
Table 4.

Impacts one year after a 1-standard-deviation exogenous decline in China’s real estate investment: Selected Commodity Prices

article image
Remark: A one-standard-deviation decline in growth is equivalent to 2-percent decline in real estate investment levels from baseline

III. Domestic Feedback

A rapid growth slowdown in real estate investment would reverberate across the economy, lowering investment in a broad range of sectors. Given strong backward linkages to other industries, especially manufacturing of construction material, metal and mineral products, machinery and equipment, a temporary, one-standard-deviation decline in real estate investment growth would cause investment in the manufacturing-heavy secondary industries to slow down by about 1½ percentage points at peak (within the first year). A slowdown in primary industry investment growth, which contains mining, is unclear. This translates approximately into a total FAI decline of about 0.8 percent from baseline level, 12 months after the shock (see Table 1).

A01ufig05

China: Peak Impact on Investment

(In percentage points, saar; 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

A01ufig06

China: Peak Impact on Exports and Imports

(In percentage points, saar; 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

Other components of demand respond in a consistent fashion. Export growth, particularly manufacturing exports, would fall by around 2¼ percentage points mainly from diminishing trading partners’ demand. The deterioration in domestic demand and weaker export growth would bring import growth down by roughly 5¾ percentage points at peak impact. Equivalently, exports and imports would fall by around 1.4 and 1.6 percent, respectively, below baseline levels, 12 months after the shock (see Table 1). A large fall in imports also reflects a significant share of processing trade in total trade. More important, the strong import responses reflect robust linkages of real estate activity to domestic industries that require inputs from abroad, namely manufacturing of construction material, mineral and metal products, as well as machinery and equipment.3 China’s REER as well as the RMB/USD exchange rate do not seem to help cushion exports in a meaningful way even though the rate of appreciation (depreciation) appears to slow down (accelerate) slightly and lasts around 2-3 quarters.

Consumption would be dampened as income and wealth expansion (including house price appreciation and stock market valuation) slows down. Real retail sales would dip by 0.2 percent below baseline 12 months after (see Table 1). The end-result would be a drop in total industrial value added and output. All in all, industrial gross value added growth would fall by around 0.4 percentage points at peak, which is consistent with around 0.3 percentage points decline in real GDP on an annualized basis.4 The impact would be felt almost immediately and would start to dissipate after 4 quarters. This would translate into a decline of about 0.3 and 0.2 percent below baseline levels for industrial value added and GDP, respectively, one year out (Table 1).

A01ufig07

China: Peak Impact on Macroeconomic Indicators

(In percentage points, saar; 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

CPI inflation would fall slightly, reflecting modest easing of price pressures as excess capacity diminishes along with demand growth.5 The overall growth slowdown is reflected in the stock market as well as labor market condition as employment growth slows in urban areas of China.

Worsened income and wealth would have important bearing on the overall and residential property markets. As demand conditions deteriorate, property market transaction volume and price would drop. For example, residential transactions volume growth would drop by around 7 percentage points at peak. One year out, residential real estate transaction volume would fall by 3 percent below baseline (see Table 1). House prices, on the other hand, would be cushioned by dwindling current and future housing supply (from shrinking housing starts). Measured using official house price statistics, which may understate residential property price inflation and deflation, house price growth would decline by around 3 percentage points at peak, or 1.5 percent below baseline 12 months after impact (Table 1). Meanwhile, the inflation in domestic prices of metal required for construction activity, such as aluminum, electrolyzed copper, and zinc would be shaved off by 1¼, 5, and 7⅓ percentage points, respectively. Deterioration in the property market climate is expected to have implications for financial institutions’ balance sheets and financial stability as well. Nevertheless, without sufficient financial indicators at monthly frequency, the model cannot uncover the relationships between a property market slowdown and financial stability indicators.6

A01ufig08

China: Peak Impact on Property Market

(In percentage points, saar; 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

IV. Global Spillover

A temporary shock to China’s real estate investment growth would have spillover implications around the world, with the impacts on G20 economies lasting approximately 4-5 quarters. In this exercise, the approximate impact on GDP growth would vary with the size of industrial production-to-GDP ratio in each economy.7 The implied peak impact on PPP-weighted G20 GDP growth is -0.2 percentage point, which translates to around 0.1 percent below baseline at 12 months after the shock originated in China (Table 2). Over all, capital goods manufacturers that have sizable direct exposure to China through exports to China in percent of own GDP and are highly integrated with the rest of the G20—therefore sharing adverse feedback from a negative shock in China with other trading partners, such as Germany, Japan, and Korea—would see more of the impact to industrial production and GDP. The results also show that global trade activity would decline (total exports and total imports for every G20 economy would weaken), which suggests that economies that derive significant benefit from global trade expansion and have deeper links via supply chain countries over the past decade, such as Germany and Japan, should be more hard hit in the second round (Table 3). Impact on Korea’s GDP peaks within the first 2 quarters and fades away more quickly, which is consistent with the fact that Korea’s direct exposure to China is large but second round effects through supply chain countries are smaller than Japan and Germany (also see Riad, Asmundson and Saito, 2012).

A01ufig09

Peak Impact on Industrial Production

(In percentage points, saar; 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

* Canada’s economic activity is represented by monthly real GDP Index, all industries.
A01ufig10

Peak Impact on Real GDP, implied

(In percentage points, saar, 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

Trade expansion with China and overall global trade would also slow as global and China demand growth weakens (Table 3). For U.K. and India, exports to China would bear the brunt of the impact, but because they are not important components of final demand in these economies, the impact on economic activity would be relatively moderate.8 Commodities exporters to China, such as Australia, Canada and Brazil, would also experience nonnegligible spillover effects on export growth.9 Australia’s relatively large direct exposure to China should imply a larger direct impact, but there seems to be other forces that blunt effect on Australia’s industrial production, for example the AUD exchange rate working as a shock absorber. Nevertheless, other indicators, such as employment growth and total import growth (not shown here), point to a slowdown in Australia’s economic activity. The impact on Indonesia’s exports would likely come through China’s coal demand. Because coal exports to China have risen sharply over the past few years, the impact on Indonesia’s output could be larger today than shown in Table 2.

A01ufig11

Peak Impact on Exports to China

(In percentage points, saar; 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

The growth spillover effects are reflected in asset prices and valuation as well. Specifically, the impact on financial wealth generation as represented by the expansion of stock market indexes in G20 economies would be tangible—by as much as 8 percentage points in Brazil and between 6-7 percentage points in Germany and Japan—and would remain for as long as 4-5 quarters. Related to this, a general decline in sovereign bond spreads (cumulative over the first 12 months after impact) seems to signal concerns about future global growth, complementing the immediate impacts on industrial production shown earlier. In the U.S.’s case, the initial flattening of the yield curve is reversed around 2 quarters after the shock, which is suggestive of the U.S.’s recovery prospects could be faster than other G20 economies. The result for Australia is consistent with the estimated impact on that country’s industrial production.

A01ufig12

Peak Impact on Stock Market Index

(In percentage points, saar; 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

A01ufig13

Peak Impact on Sovereign Bond Spreads: 10Y-2Y

(In percent; 12-month cumulative; 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

Even as nonfuel primary commodity price inflation—especially metal price inflation—retreats, the impact on global inflation appears modest. Global growth slowdown and a drop in China’s demand for base metal imports, initiated by a China real estate investment decline, could lead to a drop in iron ore, aluminum, copper, lead, nickel, and zinc price growth of between 2¾-8 percentage points. The impact on overall metal prices could last 4 quarters, with up to 5-6 quarters for lead and zinc, possibly due to weaker supply response. This is equivalent to a decline in price levels of around 1½-4½ below baseline levels, one year out (Table 4). It is unclear how crude oil prices would be affected in this exercise—the impulse responses show a drop in crude price growth, with peak at around 3 quarters after impact, but are not statistically significant.

A01ufig14

Peak Impact on World Prices

(In percentage points, saar; 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

A01ufig15

Peak Impact on World Metal and Rubber Prices

(In percentage points, saar; 1 s.d. shock)

Citation: IMF Working Papers 2012, 266; 10.5089/9781475549003.001.A001

V. Conclusion

Real estate investment account for a quarter of total fixed asset investment in China. The impact on economic activity of a hypothetical collapse in real estate investment in China is sizable, with large spillovers to a number of China’s trading partners. A 1-percent decline in China’s real estate investment would shave about 0.1 percent off China’s real GDP within the first year, with negative spillover impacts to China’s G20 trading partners that would cause global output to decline by roughly 0.06 percent from baseline. Japan, Korea, and Germany would be among the hardest hit. In that event, commodity prices, especially metal prices, could fall by as much as 0.8–2.2 percent below baseline one year after the shock.

Overall, capital goods manufacturers that have sizable direct exposure to China— especially Japan and Korea—and are highly integrated with the rest of the G20—therefore sharing adverse feedback from a negative shock in China with other trading partners—such as Germany and Japan—would experience larger decline in industrial production and GDP. Worsened global growth prospects would be reflected in asset prices and sovereign bond spreads. In that event, commodity prices, especially construction-related metal prices, would also fall.

Our sample contains at least one full cycle of real estate investment and property market in China, and represents China’s increasing integration with the world economy. Strictly from a statistical point of view, we expect a priority that this relatively short sample will make statistical relationships harder to detect and will be an important constraint on the richness of the models. Nevertheless, as the results suggest, there is still sufficient statistical information in the sample that allows us to learn something useful about China’s interaction with the world in the recent past. It is important to stress, however, that China is more important to the global economy today than our sample would suggest and a China investment bust is not likely to be a linear event as measured by the model. The impact on G20 trading partners and therefore global growth today should be larger than we report.

Appendix A: The China–G20 Macro Financial FAVAR

Why a FAVAR?

The factor-augmented vector autoregressive (FAVAR) approach offers a simple and agnostic tool to identify and measure the spillover effects of innovations in investment and real estate investment in China on various international macroeconomic, financial, trade, expectations and labor market variables. At the philosophical level, the approach works on a plausible assumption that policy makers and market participants face information constraints (similar to the econometrician) when they try to gauge economic conditions and developments, e.g. economic activity, price pressures, liquidity, and credit conditions, etc. They try to overcome these constrains by exploiting the information from a very large set of economic indicators

Technically, the approach offers a natural solution to the degrees-of-freedom problem in standard VARs by effectively conditioning VAR analysis of shocks on a large number of time series while exploiting the statistical advantages of restricting the analysis to a small number of estimated factors, which usefully summarize those time series. As it requires only a plausible identification of the shocks and not a precise identification (restriction) of the remainder of the macroeconomic model, simplicity of the VAR’s approach is retained.

By conditioning the analysis on a rich information set, the approach addresses three well-known criticisms of the low-dimensional VARs, structural VARs, and Bayesian VARs in several applications. First, it resolves the problem of mis-measurement of shocks or policy innovations—typically arising from the inability to control for information market participants or policy makers use—which leads to incorrect estimated responses of economic variables to those innovations.10 Second, it does not require the analysis to rest only on specific observable measures to represent certain economic concepts. For example, the concept of “economic activity” cannot be perfectly captured by one indicator, such as real GDP or industrial production. Including multiple indicators, e.g. retail sales and employment, could represent the concept better. “Price pressures” may be better represented by various measures of prices—CPI, PPI, commodity (metal, nonmetal, fuel or nonfuel) prices. “Interest rates” and “liquidity and credit conditions” cannot easily be represented by one or two series, but are reflected in a wider range of economic indicators.11

Finally, for the purposes of policy analysis and model validation, the impulse responses can be observed for a large set of variables that policy makers and markets care about.

The Model

Briefly, the model is a stable FAVAR in growth (except for balances and interest rates) with five common factors for each region (China and the rest of the G20 economies) and China’s real estate investment. The model uses one lag. The Cholesky factor from the residual covariance matrix is used to orthogonalize the impulses, which imposes an ordering of the variables in the VAR and treats real estate investment as exogenous in the period of shock. Specifically, the VAR ordering restricts China’s real estate investment to exogenously impact China’s common factors which then spillover onto global factors in the immediate period (one month) after the shock in a recursive fashion. By construction, there is no need to identify the factors separately because each region-specific set of common factors (or principal components) is an independent linear combination that spans the respective data set. The results are therefore robust to re-ordering within factor groups.

Formally, the FAVAR is described by a set of measurement equations (1), relating observed China data and those of the other G20 economies—the X’s, which are listed in Appendix B—to their unobserved principal components12 or factors, the C’s; and a reduced-form state equation, which governs the dynamics of the factors (2), as follows:

Xt=∧Ct+etXt*=∧*Ct*+et*(1)
[CtCt*]=[Ψ11(L)Ψ12(L)Ψ21(L)Ψ22(L)][Ct-1Ct-1*]+[utut*](2)

where * denotes the non-China factors; e’s are mean-zero error terms, which are uncorrelated with the C’s, but can be serially correlated and weakly correlated across indicators; and, finally, the u’s are reduced-form mean-zero innovations that are cross-correlated. For China, C consists of unobserved common factors (F) to be estimated as well as observed fixed asset investment or real estate fixed asset investment (R), depending on the application. These C’s should capture region-specific economic conditions or concepts that a few time series cannot represent adequately. The u’s can be written and interpreted as the sum of global exogenous shocks, driven by some global shocks and region-specific disturbances (see Boivin and Giannoni, 2008).

Equation (1) relates the information time series X to the common “forces” C, which contains unobservable factors in F and observable variables in R. It also captures the idea that both F and R can be correlated in general, representing common forces that drive the dynamics of the data, X, in each economic region.

Equation (2) is a VAR in global factors, China factors, as well as China’s real estate investment (or total investment in a different application). It specifies how these common forces evolve over time, and is usually interpreted as an atheoretic forecasting model. The off-diagonal elements of the matrix allow the shocks to affect the common factors of the other region both contemporaneously and over time. In essence, these off-diagonal matrix polynomials capture spillover effects across regions, which can be “switched on” or “off”. For instance, if the upper right element is set to zero, then the model is restricted to have no feedback to the rest of the world from China variables.

Estimation

Data are initially transformed to induce stationarity, as described in Appendix B. Then a two-step principal components approach is used to estimate the model (see Stock and Watson, 2002; and Bernanke, Boivin and Eliasz, 2005). In the first step, the common space spanned by the factors of X over time, or the C (F,R), is estimated using the first principal components of X. Denote it by Ĉ(F,R). When the number of time series is large and the number of principal components used is at least as large as the true number of factors, the principal components consistently recover the space spanned by both F and R. Since Ĉ(F,R)corresponds to arbitrary linear combination of its arguments, obtaining F requires determining the part of Ĉ(F,R) that is not spanned by R.

The second step involves estimating the FAVAR, equation (2), by standard methods with F replaced by F^. In theory, when the number of time series is large (in this case, 390) relative to the number of periods (in this case, 128), the uncertainty in the factor estimates can be ignored.

This procedure is computationally simple and imposes few distributional assumptions. This methodology provides a nonparametric way of estimating C (F,R), i.e. it does not impose the structure of a parametric model with precise distributional assumptions in the measurement equations (1).

Identification

Two distinct sets of restrictions are imposed on the system (1)-(2). The first is a minimum set of normalization restrictions on the measurement equations (1), which are needed in order to estimate the model. This is the standard normalization implicit in the principal components. The normalization is done so that solutions to the estimation problem in (1), i.e. the estimated factors F and factor loading ∧, can be distinguished from any transformation that would also satisfy equation (1), conditional on observing X. Normalization does not affect the information content of the estimated factors. The second restrictions are imposed on the factors and their coefficients in the transition equation (2) to identify the shock.

The framework then identifies unforecasted innovation in real estate investment and traces out the impact of various economic variables of interest. This framework is more appropriate for our analytical purpose than for monetary policy analysis, as the unforecasted portion of policy interest rate innovations are not interesting in the real world where central banks follow well known monetary policy rules and communicate their actions actively to influence markets.

The second set of restriction is the identification of the structural shocks in the transition equation (2). A recursive structure is assumed where all the factors entering (2) respond with a lag to change in the exogenous variable (real estate investment), ordered last. In this case, there is no need to identify the factors individually, but only the space spanned by the latent factors, F and C*. The Cholesky factor from the residual covariance matrix is used to orthogonalize the impulses, which imposes an ordering of the variables in the VAR and treats real estate investment as exogenous in the period of shock. The results are robust to re-ordering within factor groups.

As a result, no further restrictions are required in (1) and the identification of the shock can be achieved in (2) as if it were a standard VAR.

Appendix B: Data Transformation and Sources

Appendix A. Data Sources and Transformations

article image
article image
article image
article image
article image