The Effect of the Countercyclical Capital Buffer on the Stability of the Housing Market

The housing market consists of buyers and sellers, while sellers can either be households or residential construction firms. All of the agents are characterized by individual characteristics and considerations of market conditions. They trade housing units, thus forming endogenous market conditions. The housing market is competitive. No real estate seller has market power and not every potential buyer who aims at homeownership may buy one.

The financial market is populated by a set of banks. Each of them aims at profit maximization. Following their perceptions of market conditions and beliefs about future price developments, they form individual investment strategies and decide how to allocate funds. The model setting offers three investment opportunities. Banks can either hold cash, grant mortgages to potential real estate buyers or invest capital in another risky asset which is supposed to be a diversified market portfolio of financial assets that represents any alternative investment opportunities of banks. Cash earns no interest and is supposed to be risk-free. Mortgages and the alternative investment portfolio \(AI\) generate returns but are associated with default or price risk. Conducting such risky business is constrained by Basel III regulations. If financial agents decide to either grant mortgages or invest in \(AI\), they need to comply with the regulatory capital adequacy requirements \((CAR)\), including a countercyclical capital buffer \((CCyB)\).

In addition to the individual market assumptions and investment strategies, each bank is characterized by distinct balance sheet positions. These are displayed in Table 1. The composition of balance sheets is initially calibrated on the German banking sector and can be seen as a reasonable proxy for any other countries. To extract any stationary balance sheet compositions, we use the average of the last ten years which is the period from 2012–2021. Conducting business and following their investment strategies, the balance sheet positions vary every period. At the end of each period, the respective balance sheet positions are recalculated.

Cash is used to buy \(AI\) and to comply with liquidity requirements introduced by Basel III. Bought shares decrease the cash position whereas sold shares increase it. The opposite effect is realized for the \(AI\)-Portfolio. An excess in cash is mainly held by risk-averse agents. The volume of mortgage loans expands if new mortgages are granted. Repayments and defaults make the mortgage portfolio shrink.⁶ The equity position changes according to accrued gains and losses. Depending on the development of prices of the alternative investment portfolio, banks might face gains or losses. These as well as losses out of defaulted mortgages are directly translated into changes in equity. As banks have various funding opportunities, which do not directly influence the impact of the \(CCyB\), we forgo modeling those. Instead, the debt position is calculated as the difference between total assets and equity and develops passively. According to this recalculation process, it is assessed whether banks comply with or violate Basel III rules. The change in balance sheet positions is the result of individual agents’ expectations and interactions. The financial market is competitive and no single institution has market power.

The model is calibrated on empirical evidence, data obtained from the literature, or according to parameters that aim to capture real economies in terms of relations and conditions. The computational experiments are initially performed in a setting with a number of 60,000 potential buyers, 30,000 sellers, and 53 banks. Each of these agents owns a diverse set of individual features and market perceptions which ensures a high degree of heterogeneity. The initial market price of one unit of housing is set to \({P}_{h,t}=\mathrm{2,500}\) and the price development in previous periods is \(\Delta {P}_{h,t-1}=50\), and \(\Delta {P}_{h,t-2}=50\). After a sale has been conducted, the respective buyer and seller leave the market. The same holds true if a buyer was unsuccessful to buy a house for 10 periods. In this case, he is assumed to be too old to afford a future debt burden and he stays a tenant. In each period, a random number of potential buyers in a range of [60000,66666] and potential sellers in a range of [20,22] enter the housing market. Buyers’ initial debt burden rate for investing in residential property is \(r=0.12\). This is the sum of a fixed redemption rate of \({r}_{p}=0.1\) and an initial mortgage interest rate of \({r}_{t}=0.02\).¹⁴

The financial market is populated by 53 banks.¹⁵ Their initial balance sheet positions are calibrated on data obtained from BankFocus. Just like sellers and buyers, each bank is characterized by individual market perceptions and, according to this, forms its own investment strategy. In times of depreciating house prices, banks assess a risk discount to determine the collateral value of the financed property. According to German conditions, this is assumed to be 0.2 so that \(\chi =0.8\) (Bienert & Brunauer, 2007). The loan default rate for banks is set to \(D=0.01\).¹⁶ If one bank leaves the market because of insolvency, a new bank enters it with the same initial balance sheet positions but with different expectations about market developments. Following this approach, we hold the number of banks on the market fix and rule out any effects that might occur because of changing market size while at the same time ensuring agents’ heterogeneity.

The measures that determine the alternative investment \(AI\) are initialized to mimic the German stock index (DAX). The past return rate of \(AI\) is set to \({r}_{AI}=0.084\), its past fundamental value to \({f}_{t-1}^{v}=\mathrm{1,008}\). To determine \(AI{\prime}s\) fundamental value we use a drift of \(\mu =0.1215\) and an annual volatility of \({f}_{t}^{V}=0.192\).¹⁷ The parameters to initially calibrate the model are summarized in Table 2.

In the first simulation scenario, we create an experimental environment that mimics the housing market in the absence of any shocks. The economic agents enter the market, evaluate ongoing market conditions and form individual behavioral strategies. According to their utility for housing investment, buyers seek to buy a home, limited by their budget constraints. Sellers track previous market developments and decide whether to state an offer. Following the purpose to maximize profit but being constrained in business activities due to the regulatory requirements of Basel III, banks opt for financing residential property or investing in an alternative investment portfolio. The interaction of market participants creates endogenous housing market cycles. Changing market conditions feedback on the financial market and vice versa and the interplay between mortgage borrowers and banks impact the stability of the financial sector.

To analyze the effectiveness of the regulatory policies, we evaluate the dynamics of the key variables during the simulation runs. Table 3 reports the statistical values of the baseline \(CAR\)-scenario and contrasts them to those of the \(CAR+CCyB\)-scenario. The measures reveal that in both experimental environments, the house prices fluctuate cyclically around their mean. These movements are visualized in Fig. 1. The minimum price of dwellings in the \(CAR\)-scenario falls below this of the \(CAR+CCyB\)-scenario (1704,534 vs. 1741,796). This indicates that the \(CCyB\), whenever it is announced by national authorities, prevents house prices to fall down sharply. The indication is affirmed by the graphical illustration of house price movements in Fig. 1. The ongoing dynamics in the \(CAR\)-scenario push prices to deep troughs if the real estate market experiences a downturn. If prices reached a peak and depreciated in previous periods, banks progressively constrain mortgage lending. As collateral values decline, they substitute higher-risk credit businesses with higher-yielding alternative investment opportunities. The downward trend ends in a real estate price crunch. Such deeper price declines do not appear if banks only need to comply with the \(CAR\) of Basel II. Following the regulatory rule of Basel III, banks need to hold an additional amount of equity in times of excessive positive price trends. The built capital cushion shall be released in adverse market developments. Less strict equity requirements cheapen mortgage lending for banks and render it more attractive compared to investing in \(AI\). As a result, in the \(CAR+CCyB\)-scenario, banks grant mortgages more generously in times of depreciating house prices than in the \(CAR\)-scenario. The release of the \(CCyB\) ensures sufficient credit supply in downward trends. This induces prices to turn around earlier and thus averts deep price falls.

The maximum housing price is quite lower in the \(CCyB\) environment. In times of excessive credit growth, the activated \(CCyB\) is intended to stop disproportionate price increases by making lending more expensive. However, if prices are rising over time, collateral values rise likewise. Rising property values make banks less cautious and they lend more generously. Furthermore, increasing collateral values make the LTV of borrowers decline. This creates an offsetting effect. Although mortgage lending gets more expensive because of higher equity requirements, these requirements are based on the LTVs of home buyers which drop with rising collaterals. Whereas the \(CCyB\) succeeds in preventing house prices from deep price falls, it often allows prices to appreciate nearly to the same extent as with microprudential \(CAR\) only. Nevertheless, the \(CCyB\) is able to prevent very high peaks as can be seen in Fig. 1. Higher troughs and lower peeks lead to a marginally higher mean real estate price which is 2505.115.

According to macroeconomic stability, the \(CCyB\) succeeds well. The macroeconomic tool manages to reduce the standard deviation of housing prices and thus their volatility and stabilizes housing market cycles. Table 4, which displays the properties of the endogenously created housing market cycles, confirms this conclusion. The \(CAR+CCyB\)-scenario features an average cycle length of approximately 9 periods (8.9) which is lower than approximately 10 periods in the \(CAR\)-scenario. Thus, a higher number of cycles due to faster-induced turnarounds of price trends occur in the \(CAR+CCyB\)-scenario (20.0 vs. 20.9). The lower number of outbreaks (34.7) fosters the previous indication that the \(CCyB\) stabilizes housing market cycles.

In both, the \(CAR\)-scenario and the \(CAR+CCyB\)-scenario, the lowest equity a bank needs to hold is the microprudential \(CAR\)-level. In the second scenario, however, the equity requirements are extended in times of excessive house price appreciations. The average amount of equity that needs to be held in the \(CAR+CCyB\)-scenario thus exceeds this of the \(CAR\)-scenario. An increase in the cost of lending lifts mortgage interest rates to 0.038 instead of 0.036. As housing investment gets more expensive, more prosperous borrowers in terms of higher initial equity and lower LTVs will get a loan preferentially. This makes the non-default probability in the \(CAR+CCyB\)-scenario exceed that of the \(CAR\)-scenario (0.760 vs. 0.664). This apparently positive effect coincides, however, with a lower transaction rate of houses (0.179 vs. 0.185), a lower rate of homeownership (0.101 vs. 0.124), and a lower construction rate (0.115 vs. 0.179). In line with the preceding results, the average loan amount granted for residential purposes as well as the total number of mortgages is less in the \(CAR+CCyB\)-scenario than in the \(CAR\)-scenario.

To measure the solidity of the banking sector, we account for the overall Z-score of the financial market in both experimental environments. The mean Z-score in the \(CAR+CCyB\)-scenario is 2.923 and outperforms this in the \(CAR\)-scenario of 2.714. This holds true for the whole set of interacting banks. Figure 2 illustrates the distribution of the Z-scores. It reveals that not only the mean resilience of the banking sector increases which could have been caused by a few very stable institutions but also the resilience of all banks. Figure 3 shows the evolution of the mean-Z-Sore over time. In both scenarios, the development is very stable and without remarkable outbreaks. This indicates that on average financial institutions conduct business steadily in both experimental environments. Over the whole simulation period, the mean value of the Z-score in the \(CAR+CCyB\)-scenario reaches the same or a higher level than in the \(CAR\)-scenario. The higher banking soundness in the \(CAR+CCyB\)-scenario is due to two different reasons. First, the higher average equity conditions for banks equip the financial market with more loss-absorbing capital and make banks more resilient in times of economic downturns. Second, lower standard deviations of mortgage interest rates and less risky borrowers in the \(CAR+CCyB\)-scenario indicated by a higher non-default probability lower banks’ business risk.

According to the experimental results, we conclude that in dynamic market conditions and in the absence of any shocks, the \(CCyB\) is an effective macroeconomic tool as it succeeds in stabilizing the housing market and the financial market. However, it should be taken into account that despite positive impacts, it also constrains macroeconomic market activities in the housing market and restricts homeownership for borrowers with lower initial wealth or lower income. This, in turn, might negatively impact an economy’s wealth, health, and social concerns.

The first shock we investigate is a financial shock incorporated by an exogenous increase in interest rates.¹⁸ To make sure that the outcomes are comparable, we lock the random numbers for the simulation scenarios and run 10 periods in the absence of any shocks. In period 11, the shock hits the market unexpectedly. Following this approach, we ensure that the changes that occur in the housing market are only due to the shock.¹⁹

Figure 4 shows the effects of a positive interest rate shock on the development of housing prices and their means in an environment in which banks need to comply with \(CAR\) only and one in which they need to build a \(CCyB\) additionally. Since random numbers are locked for both simulation scenarios, the market develops equally to the base scenario in the absence of any shock. When the shock sets in, in both scenarios, the shock spreads to the housing market and prices are clearly affected by the rise in interest rates. The positive shock boosts mortgage interest rates. This elevates expenditures for residential property and makes housing investment less affordable. Higher lending rates, in turn, depress demand for mortgage loans. The decline in credit diminishes the demand for homeownership, dragging down house prices. These attenuation effects are amplified by banks’ borrowing constraints. In times of depreciating house prices, banks narrow loan offers (see Eq. 16) due to rising risk and the gain of the attractiveness of alternative investment opportunities. The decline in house prices depresses collateral values, furtherly tightening borrowing conditions. This financial accelerator mechanism amplifies ongoing market conditions, pushing prices into recession.

In both the \(CAR\)- and the \(CAR+CCyB\)-scenario the implications of the shock set in with a short delay. Even after the shock occurs in period 11, prices follow the usual cyclical patterns developed by the behavior of interacting agents and experience a short upturn. This mimics the sluggishness of housing prices that can also be observed in real markets. Economic agents perceive ongoing market changes and update their expectations afterwards. These processes cause the market to react within a time lag. As soon as agents’ perceptions are updated, the housing market experiences a deep bust. Due to the previously described interdependent mechanisms, prices are pushed down sharply. Although small corrections occur (Fig. 4), they are not able to reverse the ongoing depression for about 20 periods. Neither does the macroeconomic \(CCyB\) mitigate its extent. The price behavior in the micro- and macroprudential regime is almost identical. Although the regulatory authority imposes banks to hold additional capital in appreciating times before the shock hits the market, the release of the previously implied buffer is not able to mitigate the extent of the exogenous interest rate shock that hit the market unexpectedly. This reveals that the effectiveness of the macroprudential tool depends on how much buffer has been built up in the previous periods and further indicates that its effectiveness might depend on the magnitude of the shock.

After an extended period of depreciation in the housing market, the trend ends when market participants have factored in the changed market conditions. In both regimes, prices start to rise and proceed with their cyclical pattern. As mortgage lending is still expensive due to the increase in interest rates, the level of house prices falls below the base scenarios which are also portrayed in Fig. 4. Although the \(CCyB\) is not able to prevent the housing market to crash when a positive interest rate shock hits the market, it performs well in mitigating sharp price outbreaks in the aftermath of the crisis. When there is no macroprudential policy in place, housing market cycles are distinct with high peaks and deep troughs. Due to high lending costs, the demand for housing and credit fluctuates strongly. Uncertainties in market developments increase its vulnerability to crises. In the \(CAR+CCyB\)-regime in contrast, the cyclical movement of house prices is remarkably mitigated. Excessive upturns are prevented by higher capital requirements and downturns are softened by releasing them. Although the \(CCyB\) fails to prevent a housing market crisis and to mitigate the recessionary effects of a financial shock, it performs well in dampening the amplification effects of the borrowing constraint channel in the aftermath of the shock. The longer the simulation takes, the more stable the market gets in both scenarios which leads to the conclusion, that when the number of paths approaches to infinity, the average among the paths converges to a quite stable market.

Table 5 contains the statistical measures of the investigated key variables of the two regulatory regimes. The positive interest rate shock spreads to the housing market by the elevated mortgage interest rates which directly impact the demand for credit. Lowered credit volumes spill over to the demand for properties which negatively affects the price level of houses. In both the \(CAR\)- and \(CAR+CCyB\)-scenario the minimum, the maximum, and the mean price is lowered compared to the base scenarios. As in the environment without any shocks, the minimum value of the \(CAR\)-scenario which is 833.005 falls below that of the \(CAR+CCyB\)-scenario (1057.514) whereas maximum values are quite equal (407.500 vs. 3232.250). Although the price development is identical before the shock hits the market and during the recession, the \(CAR+CCyB\)-scenario outperforms the \(CAR\)-scenario in terms of housing market stability. This is indicated by a lower standard deviation in the \(CAR+CCyB\)-scenario that amounts 420.755. Due to the lowered house price fluctuations in the aftermath of the crisis, the \(CCyB\) is effective in stabilizing the housing market overall.

As mortgage interest rates are the channel through which the shock on the financial market spills over to the housing market, they are significantly increased. As in the base scenario, the mortgage interest rates with \(CCyB\) are slightly higher (0.084 vs. 0.090). This is in line with the increased capital requirements during the shock recovery phase that prevent excessive credit growth. Through the inverse relationship of the mortgage-to-income ratio and the non-default probability of borrowers (see “Mortgage Supply”), higher housing expenditures amplify the riskiness of credit exposures and sharply decrease borrowers’ probability of not defaulting. As higher capital requirements tighten banks’ business activities, banks will lend to customers with higher LTVs in the \(CAR+CCyB\)-regime which lifts the non-default probability to 0.291 compared to 0.272 the \(CAR\)-regime.

Similar to the base scenario, the lifted capital requirements in the \(CAR+CCyB\)-scenario affect activities in the housing market restrictively and limit transaction rate, homeownership rate, and construction rate to a lower level than without \(CCyB\). In the \(CAR+CCyB\)-scenario, the mean transaction rate is 0.092, the homeownership rate 0.098, and the construction rate 0.056 while in the \(CAR\)-scenario they reach 0.180, 0.177, and 0.069. Compared to the base environment, the ratios are significantly reduced due to higher lending costs. The loan amount as well as the number of accepted credit exposures is significantly smaller in the shock environment than in the base environment and the values of the \(CAR+CCyB\)-regime fall below the \(CAR\)-regime.

Whereas the housing market is hit by an exogenous interest rate shock that generates high volatility in housing prices, the Z-scores of both regulatory schemes exceed this of the base scenarios, indicating higher banking soundness. Increased interest rates raise financial institutions’ profits. These profits render the financial market a profound equity base that increases its solidity. Considering the low values of the non-default probability of borrowers, the assumed effect on banks’ stability would be the other way around. The standard deviation of mortgage interest rates reveals, however, that profit out of mortgage lending is nearly as stable as in the base scenarios. This fact indicates that in an environment with a positive interest rate shock, banks, even more, constrain lending to borrowers with initial wealth or favorable LTV ratios. The impact of those loan-granting policies on the rest of the population is indicated by a decreasing transaction, homeownership, and construction rate in the shock scenario. Figure 5 displays the mean Z-scores of both regulatory regimes. As in the base scenarios, the \(CAR+CCyB\)-regime achieves higher values for the interactive banks and achieve a mean value of 3.116 vs. 2.862 in the \(CAR\)-scenario. The development of the mean Z-score over time (Fig. 6) reveals that banking soundness is only slightly affected by the positive interest rate shock. In the aftermath of the shock, the values remain quite stable in both the \(CAR\)- and the \(CAR+CCyB\)-regime whereas the second one exceeds this of the first one steadily. The high values of Z-scores in this shock scenario indicate that only accounting for microprudential measures is not sufficient to evaluate the impact of occurrences on the overall economy. Instead, macroeconomic effects need to be investigated that can be effectively lowered by macroprudential policies.

Relating to overall stability measures, the \(CCyB\) is effective in stabilizing the housing as well as the financial market in the aftermath of an exogenous positive interest rate shock. These results are verified by Table 6 which shows the properties of cycles under the \(CAR\)- and the \(CAR+CCyB\)-regime if a positive interest rate shock hits the market unexpectedly. Under the \(CAR+CCyB\)-regime, the average length of housing market cycles is lowered to 9.0 due to earlier initiated turnarounds and the number of outbreaks is reduced from 79.0 to 51.0.

The second shock we investigate is a positive housing demand shock incorporated as an increase in preferences for housing on the demand side.²⁰ Similar to the financial shock examined in the previous chapter, the housing demand shock is unanticipated by market participants and hits the market unexpectedly in period 11. The periods before the shock sets in are hold fix.

The shock generates expansionary effects in the housing market. The increase in the weight of residential property in the utility function for borrowers fuels the demand for housing. This pushes prices upwards and leads to high minimum, maximum, and mean prices under both regulatory regimes. In the \(CAR\)-scenario, the mean price level reaches 3247.530 which is slightly lower than this of the \(CAR+CCyB\)-scenario (3261.565). The supply still develops endogenously by the perceptions and expectations of the economic agents. The increased demand cannot fully be absorbed by the housing supply, generating a sellers’ market and driving house prices upwards. Figure 7 contrasts the development of dwelling prices and their means of the \(CAR\)- and \(CAR+CCyB\)-regime in the demand shock environment and compares them to the baseline scenarios. Until the shock hits the market in period 11, in both regimes, the price developments are similar to the base scenario. As soon as potential house buyers put more emphasis on being a homeowner, prices drift upwards, experiencing high positive peaks during the simulation, especially in the \(CAR\)-only regime. Positive price trends reverse sharply to the upper price level of the baseline scenario. If banks only need to comply with microprudential regulatory measures, the house price cycles are distinct. Price appreciations lead to high peaks which are followed by recessionary periods that end in deep busts. The extent of the fluctuations strongly corresponds to those of the \(CAR\) base scenario (see Table 7). In the environment where the \(CCyB\) is installed, house price movements are flattened. The macroeconomic tool mitigates excessive price increases and softens downturns also to a similar extent as it does in the \(CAR+CCyB\) base scenario (Table 7). These results show that a positive housing demand shock directly affects housing prices, leading them to a higher level. However, compared to the baseline scenarios, the stability indicated by the standard deviation of housing prices in Table 7 amounts 450.363 in the \(CAR\)-scenario and 359.012 in the \(CAR+CCyB\)-scenario and thus, is not negatively impacted.

Whereas the reactions of the housing market to the financial shock set in with a delay, the housing demand shock affects prices immediately. This reveals that the transmission time and the impact of the shock are addicted to its origin.

The increased demand for buying a residential property that fuels house prices spill over to the financial market through greater demand for credit. The mean mortgage volume granted per period of the \(CAR\)- and the \(CAR+CCyB\)-regime strongly exceeds this of the base scenarios and the financial shock scenarios (see Table 7). The number of granted loans, in contrast, is comparable to those of the base scenarios. This reveals that the increase in mortgage volume is induced by the rise in house prices and mimics one spillover effect from the housing market to the financial market. Rising house prices further impact mortgage lending as they soften the borrowing constraints for potential buyers. Higher collateral values allow borrowers to extend credit volume which in turn further drifts up prices, launching the financial accelerator mechanism and reinforcing ongoing market developments. The dampening effects of the \(CCyB\) on these market dynamics are reflected in the loan amount (341,900.17) and the number of accepted mortgages (1824). Both measures are below those of the \(CAR\)-regime. Forcing banks to hold additional capital mitigates excessive credit growth and attenuates self-reinforcing effects that arise from the interplay of the housing and the financial market.

As a response to excessive demand for credit, banks raise their mortgage interest rates in comparison to the base scenarios (Table 7). Since the introduction of the \(CCyB\) makes mortgage lending more expensive for banks, it has an additional increasing effect on interest rates. This cost-increasing effect of mortgage borrowing restrains the demand for housing and counteracts excessive house price appreciations. Higher collateral values counterbalance higher expenditures for housing investment, letting the non-default probability of borrowers decrease in comparison to the base scenarios. The need for additional capital in the \(CAR+CCyB\)-environment counteracts this effect and leads to an outperforming non-default probability of 0.596 compared to 0.560. The extended capital requirement has a tightening effect on mortgage granting and makes it more secure. It is able to cushion future probable losses and makes the probability of not defaulting increase compared to the \(CAR\)-environment.

The transaction rate and the construction rate exceed those of the base scenarios and the financial shock scenarios. As the existing supply of houses is not able to satisfy the extensive demand, construction firms extend their business activities and build new dwellings. Both variables are slightly depressed in the \(CAR+CCyB\)-scenario as the macroprudential measure has a restrictive effect on market activities. The homeownership rate of 0.056 in the \(CAR\)-scenario and 0.074 in the \(CAR+CCyB\)-scenario falls below this of the scenarios described in the previous chapters. Due to the positive housing demand shock that induces rising house prices, the number of potential home buyers rises excessively whereas the supply of dwellings is limited. A huge amount of unsatisfied housing demand diminishes the homeownership rate. As the \(CCyB\) has a limiting effect on house price appreciations, it positively affects the rate of homeowners and generates a higher value than in the \(CAR\)-environment.

The Z-score of the two regulatory schemes in a market that experiences a positive housing demand shock is lower than in the respective baseline and financial shock scenarios. It amounts to 2.432 under the \(CAR\)-regime and 2.870 under the \(CAR+CCyB\)-regime. This indicates that an external shock affects the soundness of the financial market more than a shock that has its origin in the financial market. And though, as in the previous simulation scenarios, the \(CCyB\) performs well in stabilizing financial institutions in the event of a positive housing demand shock and makes them more resistant to potential crises. This can also be seen in Fig. 8 which visualizes the mean Z-scores of the interacting banks in the \(CAR\)- and the \(CAR+CCyB\)-regime. Figure 9 shows the development of the Z-score over time. The Z-scores in the \(CAR\)-environment fluctuate more strongly than in the \(CAR+CCyB\)-environment. These variations also exceed those in the baseline and the financial shock scenarios.

Table 8 summarizes the cycle properties of both regulatory schemes in the event of a positive housing demand shock. The measures support previous results and reveal that the macroprudential \(CCyB\) prevents excessive housing market cycles. It lowers ongoing price appreciations or depreciations and induces market trends to turn earlier. This is revealed by a lower average cycle length (6.8 vs. 10.4) which is accompanied by a higher number of cycles. Fewer outbreaks in the \(CAR+CCyB\)-regime (29.0 vs. 39.0) indicate that the cycles oscillate closer around their mean value.

As the simulations show, the \(CCyB\) is an effective tool to mitigate excessive house price increases and to moderate spillover effects to the financial market in the event of a positive housing demand shock. Forcing banks to hold additional capital in times of excessive credit growth dampens house price appreciations. The smoother development of housing prices is transmitted to the rest of the economy more softly and thus, the \(CCyB\) moderates the impact of a positive housing demand shock on the macroeconomy.

The perceptions and expectations of economic agents are decisive key indicators for the development of the housing market. In this section, we model a housing bubble similar to that which caused the latest financial and economic crisis and evaluate the impact of agents’ distinct expectations about future housing demand appreciations on the key variables of the housing and the financial market. In this model environment, agents expect housing demand to rise at a certain point of time in the future. This expectation may be driven by different macroeconomic developments, such as a positive news shock about housing productivity, or housing supply, monetary policy that indicates a decline in future interest rates, or an upcoming subsidy for homeownership.

In contrast to the financial shock and the housing demand shock described in the previous sections, in this model environment, the agents anticipate a future change in market conditions. As a result, the investigated shock does not have a fundamental basis but is expectation driven by market participants. To simulate this, we assume that potential home buyers expect an increase in demand to occur in period 20. In period 20, however, the expectation does not materialize.²¹

The positive anticipations of the interacting agents generate expansionary effects in the housing market that are visualized in Fig. 10. The expectations of a future increase in housing demand fuel prices immediately. Potential home buyers’ assumptions of a higher demand for residential property are associated with rising expenditure for housing investment in the future and suppressed supply. As a result, potential buyers invest in housing today for speculative purposes, causing current demand to rise and inducing inflationary effects on prices. As in “Positive Demand Shock”, these market mechanisms spill over to the financial market and fuel mortgage lending as well as mortgage interest rates. Similar to the positive demand shock scenario (“Positive Demand Shock”), house prices react without any delay. This supports the hypothesis that those markets in which a shock originates record the first effects as transmission mechanisms are short. The expansionary effects appreciate house prices during the expectation period to a higher level compared to the base scenarios. The optimism about future housing demand generates a housing bubble in both the \(CAR\)- and the \(CAR+CCyB\)-scenario inducing a macroeconomic boom characterized by a co-movement between excessive residential investment, house price increases, and granted mortgage loans. During the simulation, the price level of the \(CAR+CCyB\)-scenario is mostly lower than this of the \(CAR\)-scenario. This indicates the \(CCyB{\prime}s\) effectiveness in mitigating excessive price increases. However, it is not successful in preventing a housing bubble to occur. In the \(CAR+CCyB\)-scenario, expectations of a future increase in housing demand push current prices upwards nearly to the same extent as in the \(CAR\)-scenario. When the anticipated demand increase does not materialize in period 20, in both investigated regulatory regimes prices follow their ongoing trend for a few more periods and fall into recession afterward. The failed expectation makes the housing bubble burst and causes house prices to collapse. The previous self-reinforcing expansionary effects reverse and impact the macroeconomy oppositely. The release of the \(CCyB\) in times of economic downturns only cushions the huge price decline marginally.

The impact of a housing bubble induced by an expected increase in housing demand in the future is more distinct than an unanticipated positive interest rate shock (“Financial Shock” ) and an unanticipated positive housing demand shock (“Positive Demand Shock”). The price drop that follows a housing bubble is deep and it takes long for the housing market to recover. After an extended period of price depreciation, the market turns, and prices slowly approach the price level of the base scenarios. In the case of unanticipated shocks, economic agents price in changing environments earlier. Expectations about future market conditions distort the investment plans of potential home buyers, price levels, and mortgage borrowing levels that suddenly reverse when the expectations do not realize. The unrealized anticipations cause a dramatic drop in the housing market and aggregate variables fall below their initial level. During the recovery phase, the macroprudential \(CCyB\) affects house price movements positively. It induces prices to turn around earlier ending the recession at a higher level. In the aftermath of the burst housing bubble, house price movements are more stable and less distinct in the \(CAR+CCyB\)-scenario. Similar to the financial shock environment, the \(CCyB\) is not able to prevent the crisis caused by an anticipated increase in housing demand but it is successful in stabilizing housing markets in the aftermath of the crisis.

Table 9 contains the investigated key variables in the housing bubble environment. The presented measures summarize the occurrences in the housing and financial market which can be classified into three parts: an excessive boom during the expectation phase, a recession, after the housing bubble burst, and a recovery phase when the recession is overcome. The house price level indicated by the mean of all simulation periods amounts 2479.766 without \(CCyB\) and 2315.595 with \(CCyB\) and thus is lower than in the base scenario. The excessive price increases during the expectation periods are compensated by the following price drops in the recession. In this environment, the price level of the \(CAR+CCyB\)-regime falls slightly behind this of the \(CAR\)-regime as price outbreaks are mitigated by the additional capital requirement, and busts are flattened by its relief. The high standard deviation in both the \(CAR\)- and the \(CAR+CCyB\)-scenario (595.432 and 580.621) indicates the instability of the housing market if market conditions are disturbed by a housing bubble. The coefficients exceed the fluctuations measured in the other shock scenarios. Although the macroeconomic buffer is not able to prevent a crisis caused by a housing bubble, it succeeds in extenuating its extent. The lower standard deviation of housing prices in the \(CAR+CCyB\)-scenario verifies the visualization of Fig. 10. House price oscillations are flattened and cycles are less distinct if the macroprudential tool is installed. When the aftermath of the shock is overcome, prices get very stable and hover slightly around their means.

As the three parts of the housing bubble environment partially offset market mechanisms from previous periods, the mean mortgage interest rate is quite similar to that in the base environment. The increased demand during the expectation phase transmits across the economy in a qualitatively similar manner as in the positive demand shock (“Positive Demand Shock”) and banks raise their lending rates. If banks need to hold additional regulatory capital, mortgage lending is more expensive which transmits to higher lending rates of 0.045 in the \(CAR+CCyB\)-regime. An offsetting effect sets in as soon as prices decline.

The compensating effects of market phases are also reflected in the non-default probability of borrowers. During the boom phase of the housing bubble environment, the probability of not defaulting is comparable to this of the housing demand shock environment. In the recession, it decreases heavily. When the market recovers, it recovers as well. As a result, the non-default probability in the housing bubble environment is lower than the values of the base scenarios. Similar to the base environment and the other shock environments, the \(CCyB\) lowers default risk. The probability of not defaulting amounts 0.654 whereas this without an installed \(CCyB\) reaches 0.555.

The housing bubble environment reaches the second lowest transaction rate (0.171 for the \(CAR\)-regime and 0.153 for the \(CAR+CCyB\)-regime). The extended recession and the slow recovery of the market limit trading. The same holds true for the construction rates in both regulatory regimes. The pronounced price drops discourage construction firms from building first-time occupancies. As the \(CCyB\) impacts market activities restrictively, the ratios are depressed in the \(CAR+CCyB\)-regime compared to the \(CAR\)-regime. The same effect can be seen in the homeownership rate which is lower if banks need to comply with the \(CCyB\)-regulation.

Considering all simulation periods, the mean values of the loan amount in both the \(CAR\)-scenario and the \(CAR+CCyB\)-scenario are higher than in the base scenarios which is induced by the excessive boom phase during the first 20 periods. As identified in the other simulation scenarios and in line with the measures so far in the housing bubble environment, the amount of granted loans in the \(CAR+CCyB\)-scenario lies behind the \(CAR\)-scenario (184,431.38 vs. 188,175.44). The same holds true for the number of accepted mortgages (2001 vs.2093).

The stability of the financial market, indicated by the Z-score can be slightly increased if banks need to comply with the \(CCyB\) and reaches a value of 2.541. As in the previous environments, the additional capital buffer creates a solid equity base and makes financial institutions less prone to crises. Figure 11 illustrates the distribution of Z-scores within the financial market between the interacting banks. Under both regulatory regimes, some banks feature quite high Z-scores which impact the mean Z-score of the financial market and thus the overall stability positively. As to be seen in Fig. 12, the price appreciations during the expectation phase pushed Z-scores upwards. With ongoing market developments, these appreciations are reversed. Similar to the housing demand shock environment, the development of the Z-scores over time is less stable under the \(CAR\)-regime than under the \(CAR+CCyB\)-regime.

The stabilizing effects of the \(CCyB\) are verified in Table 10. The length of housing market cycles is lowered to 20.3, the number of cycles increases (3.4), and the number of cycle outbreaks that exceed the standard deviation decreases to 32.0 compared to 51.0. Thus, the simulations show that in the event of a housing bubble, the \(CCyB\) is not a prevention but a helpful tool to reduce spillover effects to adjacent markets and to stabilize the economy. Although house price appreciations are dampened during the boom phase, it goes on for too long even if the macroprudential feature is installed. The recession that follows the burst cannot be avoided, its extent can only slightly be mitigated. Overall, however, the \(CCyB\) helps to attenuate the impacts of a housing bubble on the housing and the financial markets.