The Impact of COVID-19 on Real Estate Markets in Germany

4.1 Rents and House Prices

We construct house price and rental indices using a comprehensive and detailed dataset on housing RWI-GEO-RED (RWI and ImmobilienScout24 2022a, b). The dataset includes a wide range of property characteristics, enabling us to calculate hedonic indices.

To create a quality-adjusted index, we employ panel hedonic regression as follows:

where h, i, j, and t index properties, zip codes, municipalities, and times (i.e. month-year), respectively. The variable p denotes rent or price in euros per m², x hijt ′ includes a set of property characteristics, and α _ijt denotes zip code-municipality-time fixed effects.^[4] Our zip code-level quality-adjusted prices (or rents) are the estimates of the fixed effects in Equation (3): α ̂ ijt = log ⁡ p ̂ hijt − x hijt ′ β ̂ .

4.2 Zip Codes, Labor Market Regions (LMRs), and Central Business Districts (CBDs)

Our zip code data come from the Postleitzahlen Deutschland and covers all zip codes in Germany. The data contains information on zip code boundaries and associated towns and cities.

Kosfeld and Werner (2012) define 141 labor market regions (LMRs) in Germany, based on commuting flows. The delineation of these regions is based on combining one or more administrative regions at the county level to create self-contained labor markets. The boundaries of local labor markets are defined so that commuting flows within labor market regions are relatively large compared to commuting between regions, with an upper limit of 45–60 min for commuting time.

We use these LMRs to define the CBDs as the geographical center of each LMR, weighted by the number of inhabitants at the 1 × 1 km grid-cell level.^[5] The information on the number of inhabitants comes from RWI and Microm (2022). Figure 1 shows the constructed CBD locations and the respective distance between the CBD and the center of each zip code.

Figure 1:

CBD locations and distance to the CBD. Notes: The distance is calculated as the euclidean distance between the geographical center of the zip code and the CBD (dots) based on the LMR definition.

For a categorical classification of CBD and suburb, we rely on a data-driven definition of groups by using percentiles of the distance to the CBD. Locations within the 10th percentile are considered to belong to the CBD area (P _10 % ≈ 6.1 km). We classify locations above this threshold but below the median distance as suburban areas. For a more detailed analysis, we split the suburban areas into two rings: ring 1 with distances up P _30 % ≈ 12.4 km; ring 2 with distances up P _50 % ≈ 18.3 km. Zip codes with larger distances to the CBD lay outside the core area and are not further considered in this classification.

4.3 Amenities

We distinguish between two types of amenities – consumption amenities and environmental amenities. To model the presence of amenities in urban areas accurately, we rely on OpenStreetMap (OSM) data, which we extract for the period 2017 to 2021.^[6] Consumption amenities combine places to eat, places for education, and places for entertainment. We categorize green spaces and water bodies as environmental amenities.^[7]

The consumption amenities data are collected as point data. It is possible that we list multiple buildings belonging to an amenity. For example, even though a city may have only one university, its buildings may be scattered throughout the city, resulting in multiple data points in our sample. The environmental amenities data are collected as polygon data.

Figure 2 shows the distribution of consumption amenities (left panel) and environmental amenities (right panel) relative to the distance to the CBD for all zip codes. We report consumption amenities based on frequency, as these are represented by points in space. Environmental amenities are reported in terms of the area covered relative to the size of the zip code, as these are represented as polygons. The figure shows that consumption amenities have a higher frequency closer to the CBD, as indicated by the decreasing slope of the trend line. The pattern is less clear for environmental amenities. Here, the slope of the trend line is close to zero, indicating evenly distributed amenities. However, the stylized facts (Section A.2) show locations further away from the CBD benefit, on average, from larger environmental areas. The trade-off between consumption and environmental amenities may influence households’ decisions on where to live. Those prioritizing consumption amenities may reside closer to the CBD, while those who value environmental amenities may prefer suburban locations.

Figure 2:

The gradient of amenities (raw values) as a function of the distance to the CBD. Notes: Both panels illustrate the relationship between amenities and their distance to the CBD, with data points coded green representing zip code values and darker points representing averages within a 2 km distance bins. (a) Consumption amenities. (b) Environmental amenities.

We transform the raw data into a zip code dataset to make consumption and environmental amenities usable in the analysis. We use the frequency of consumption amenities and the area covered by environmental amenities, which are normalized into an index using z-score normalization.^[8] Following this strategy has the advantage that the resulting indices have a mean of zero and a standard deviation of one, making them comparable despite having different units.^[9]

5.1 Development of Bid-Rent Curves

While the pandemic appears to be affecting the trend of apartment rents in Germany, average house prices have remained unaffected, as shown in Figure 3. In Germany, average rents started decreasing at the beginning of 2020, coinciding with the onset of the pandemic. Rents recovered to the pre-pandemic (late 2019) levels in late 2020. However, the trend repeats starting in January 2021, with rents falling throughout 2021, albeit at a lower rate, paralleling the second wave of the pandemic. Since increasing rents by landlords is highly unlikely due to rental controls in Germany, the downturn in rents may have been triggered by the decline in demand, especially in major German cities and high-density areas.

Figure 3:

Trends of house prices and rents relative to March 2020. Notes: Note that the data is smoothed using a 3-month moving average.

With respect to house prices, the figure shows a steady increasing pattern which is not interrupted by the pandemic. This may be due to several factors. House prices are forward-looking, and home buyers likely anticipate a recovery from the pandemic by 2021 therefore, showing no (immediate) response to COVID-19. In addition, low interest rates may have supported demand for houses, and housing supply constraints on new construction may have limited the availability of homes for purchase.

Overall, the trends suggest that while the pandemic directly impacted apartment rents, house prices were more resilient, likely due to a combination of demand, supply, and financial factors.

A key finding in the literature is that suburban housing prices and rents increased while those in the CBD locations decreased during the pandemic in the US (see, for example, Gupta et al. (2022)). However, according to our results, this pattern is not strongly observed in the German housing market. Figure 4 shows the relationship between hedonic prices (left panel) and hedonic rents (right panel) and the distance to the CBD for March 2020 and March 2021. Both price and rent gradients have a negative slope, consistent with the key prediction of the AMM model that prices or rents decrease with the distance to the CBD.

Figure 4:

The price and rent gradients for March 2020 and 2021. Notes: The figure illustrates the relationship between the distance to the CBD, house prices, and apartment rents, comparing the pre-pandemic (March 2020) and pandemic (March 2021) periods. As expected, in both periods, the gradients are negatively sloping, indicating that prices and rents decrease as the distance from the CBD increases. Lighter points represent zip code values, whereas darker points represent averages within a 2 km distance bins. (a) Prices, (b) Rents.

However, we do not observe any effects of the pandemic leveling the bid-rent curves, at least not in terms of prices. The price curves appear equally parallel in March 2020 and March 2021 (Figure 4 left panel). The slope is slightly flatter for the rent gradient in March 2021 than in March 2020, although the difference is marginal. We also provide additional evidence that strengthens this finding in the Appendix by plotting the change in house prices and rents against the distance to the CBD and comparing the pandemic levels with pre-pandemic values (see Figures A.3 and A.4). If the US pattern carries over to the German setting, we expect suburban areas to experience positive growth in both prices and rents as suburban housing or rental demand increases due to the pandemic. However, there is no clear evidence that this is the case for the German housing market.

Figure 5 displays the estimates of the slopes of the gradients specified in Equation (1) for each month. The slope of the price gradient shows greater fluctuations in the point estimates but maintains a similar trend as the pre-pandemic time. In contrast, the slope of the rent gradient exhibits a smoother gradual increase over time, starting in late 2019, suggesting that the rent gradient has become flatter during the pandemic.

Figure 5:

Estimates for the price and rent gradients. Notes: This figure shows the estimated slopes of the price and rent gradients, which were obtained from the regression in Equation (1). log  p and log  r are regressed on log  dist × time, with time and LMR fixed effects. The zip code controls include the logarithm of purchasing power, the share of males aged 18–45, and the share of individuals with a German background. (a) Prices, (b) Rents.

Quantitatively, our estimation results show that the estimated slopes of the gradients are: ( δ ̂ t , δ ̂ t + 1 ) = ( − 0.052 , − 0.047 ) ; ( − 0.044 , − 0.034 ) , where t = March 2020, t + 1 = March 2021, for prices and rents, respectively. See Table A.3 for the full set of estimates.

That means, from March 2020 to March 2021, the slopes of the gradients have changed by: Δ δ ̂ = δ ̂ t + 1 − δ ̂ t = ( 0.005 , 0.010 ) for prices and rents, respectively. The positive changes in the slopes show the flattening of the gradients. However, these values are economically insignificant to suggest that the pandemic has caused the same clear flattening of gradients in Germany as observed in the US. For the US, Gupta et al. (2022) found the change in slope for price ( δ ̂ = 0.013 ) , and the slope change for rents ( δ ̂ = 0.0319 ) , from December 2019 to December 2020.^[10]

Overall, the changes in slope estimates during the pandemic appear small: the decreases in the gradients’ slopes are 9.38 % and 22.40 % for house prices and rents, respectively. This might suggest that the pandemic has led to a flattening of gradients in Germany, but not as clearly and strongly as observed in the US.

Another approach, as used by Ramani and Bloom (2021), for identifying the donut effect involves dividing the data into CBD and suburban regions and examining the growth of house prices and rents. Figure 6 shows the evolution of prices for the CBD (in red) and the two suburban rings (in green/blue) estimated for the full sample. The plot shows that the evolution of rents breaks with the onset of the pandemic (right panel Figure 6). While rents increased before March 2020, they stagnated shortly after the start of the pandemic. Rents increased again in the summer of 2020 but then decreased for all three groups. At the beginning of 2021, rents in CBDs appear to fall more sharply than in locations further away from the city center. This suggests a delayed donut effect, at least partially and less strongly than in the US.

Figure 6:

The donut effect for the full sample. Notes: The vertical dotted line indicates March 2020, the start of the pandemic. (a) Prices, (b) Rents.

The evolution for prices is even less pronounced (left panel Figure 6). Here, all three groups move in the same direction and show a continuous increase over time.

The donut effect is most prominent in big cities (Ramani and Bloom 2021). We test this prediction by re-estimating the model only for the 15 largest German cities with approximately 500,000 residents or more. Figure 7 shows that all lines are more volatile for both rents (right panel) and prices (left panel) across all areas. The partial donut effect for rents from the full sample also disappears. Thus, we cannot confirm that the donut effect is strongest in the largest cities in Germany. The indicated donut effect in the full sample comes from medium-sized cities with a population of at least 100,000 but less than 500,000, as shown in Figure A.5 in the Appendix. We also do the same exercise for small cities (below 100,000 in population) and find no evidence of the donut effect (see Figure A.6 in the Appendix).

Figure 7:

Donut effect for the 15 largest cities. Notes: The vertical dotted line indicates March 2020, the start of the pandemic. (a) Prices, (b) Rents.

5.2 The Role of Amenities

The literature identifies WFH and amenities as potential channels to explain the housing market results. Due to data limitations and since WFH seems to be weakly valued in Germany (see, for example, Nagler, Rincke, and Winkler (2022)), we focus on the second channel – the change in the valuation of (urban) amenities.

To assess the valuation of amenities, particularly the potential change during the pandemic, we estimate Equation (2), which includes a continuous time measure with effects relative to March 2020. Figure 8 shows the plotted interaction effects for each month for prices and Figure 9 shows the effects for rents without spatially restricting the sample. Regardless of the type of dwelling and the amenity considered, there seems to be a slight decreasing trend in the value of amenities after the pandemic, but almost all interactions are insignificant. There is little to no evidence that the value of amenities changed during the pandemic. We have also tested the same analysis for CBD and suburbs in the Appendix, with the same result (Figures A.8–13).

Figure 8:

Amenities’ values over time with respect to prices – full sample. Notes: The vertical dotted line indicates March 2020, the start of the pandemic. (a) Consumption amenities, (b) Environmental amenities.

Figure 9:

Amenities’ values over time with respect to rents – full sample. Notes: The vertical dotted line indicates March 2020, the start of the pandemic. (a) Consumption amenities, (b) Environmental amenities.

Overall, the results suggest that bid-rent curves in Germany have not flattened, except for rents, where there is only slight evidence. This conclusion is reasonable because WFH is not seen as a compelling job attribute in Germany, the value of amenities remains unchanged, and there is little population movement during the pandemic (Figure A.7).

5.3 Discussion on Differences Between the US and Germany

The disparities in the pandemic’s donut effects on the US and German housing markets can be attributed to institutional and cultural differences between the two markets.

Homeownership Culture and Rental Market Dominance. The United States and Germany exhibit contrasting housing market dynamics. The US is characterized by a strong homeownership culture, whereas Germany is dominated by the rental market. In the United States, the rate of homeownership has remained relatively stable at around 63–65 % over several decades. This reflects both a cultural preference for home ownership and the presence of supportive policies that have proven resilient in the face of global crises (see Federal Reserve Bank of St. Louis (2024)). Surveys conducted by the National Association of Realtors (NAR),^[11] have consistently demonstrated that the primary motivation for purchasing a home is the desire to own property. In 2019 and 2020, 29 % of all buyers identified this as their main reason (National Association of Realtors (NAR) 2019, 2020).^[12] The United States’ tax policies, such as the Home Mortgage Interest Deduction (allowing tax deductions on the first $750,000 of home mortgage interest), provide an incentive for home ownership by reducing the tax burden associated with mortgages, thus making home buying more attractive and financially feasible for many (see Internal Revenue Service (IRS) (2023)). Conversely, in Germany, only 41.8 % of households owned their homes in 2022, a figure that serves to illustrate the extent to which rental culture has become ingrained in German society (see Federal Statistical Office (2024)). Despite the financial stability, many Germans continue to prefer renting. The Bundesbank Panel Survey of Households in 2022 revealed that only 37 % of German households would choose to buy a home even if no down payment were required (see Deutsche Bundesbank (2023)). Germany’s rigorous tax policies on home sales, which discourage hasty resales, and robust rental regulations contribute to a diminished sense of urgency to purchase, further dampened by a dearth of flexibility in relocating. This structure fosters stability but stands in stark contrast to the US market, where both culture and policy support a highly mobile, ownership-oriented housing market, leading to greater flexibility and contributing to differences in the estimated effects.

Government intervention and market stabilizers The German and American governments have adopted markedly different approaches to housing market stabilization, reflecting distinct priorities and regulatory philosophies. In Germany, government institutions play an active role in maintaining housing affordability and stability through the implementation of substantial rental regulations, the provision of direct housing benefits, and investments in social housing, which is also known as “sozialer Wohnungsbau”. The “Wohngeld” (housing benefits) program, which was initiated in 1965, provides financial assistance to both renters and owner-occupants, thereby ensuring that housing costs remain within the financial capabilities of lower-income households. In 2020, approximately 1.5 % of German households benefited from this support, as documented by the Federal Ministry for Housing, Urban Development and Building (BMWSB) (2023a) and Federal Ministry of Justice (FMJ) (2008). Furthermore, Germany has pledged approximately one billion euros annually from 2020 to 2024 to subsidize social housing construction, thereby facilitating access to affordable housing for low-income households through low-interest credit programs (see Federal Ministry for Housing, Urban Development and Building (BMWSB) (2023b)). The Tenancy Law Amendment Act (TLAA) of 2013 and the Rent Freeze (Mietpreisbremse) of 2015, two legislative tools, serve to further cap rent increases in high-demand areas. This has the effect of preventing significant price surges and promoting long-term housing stability.^[13] In contrast, the United States relies more on market-driven forces to dictate housing costs, with limited government intervention primarily in the form of targeted support, such as the Housing Choice Voucher Program (Section 8), which aids low-income households in renting private market properties (see NYC Housing Authority (2024)). This market-based approach offers greater price flexibility, enabling more rapid adjustments in response to shifts in demand, as evidenced by the rising prices of housing in suburban areas in response to increased demand. Although this enables dynamic pricing and relocation, it frequently lacks the structural stabilisers present in Germany’s housing market, resulting in greater volatility in housing costs.

Housing supply constrains and zoning differences The discrepancy in urban and suburban development between Germany and the United States can also be attributed to the interplay of housing supply constraints and zoning regulations. In Germany, the combination of restrictive policies and a constrained housing supply has led to a notable limitation in suburban expansion. This is evidenced by the decline in construction permits issued in recent years, which has further constrained the availability of housing.^[14] The limited availability of housing, coupled with rent controls, has resulted in a relatively stable demand for housing across urban and suburban areas. This has led to an increase in the length of time residents are willing to remain in their current place of residence. The German zoning policies that prioritize a mixed-use approach permit residential areas to exist in close proximity to commercial and business spaces, thereby creating neighborhoods where housing is situated in close proximity to essential services and amenities (see Hirt (2007)). In contrast, the United States employs a single-use zoning approach, whereby extensive areas are designated for exclusive use as residential, commercial, or business zones. This approach frequently restricts the construction of multi-unit buildings in single-family home neighborhoods, resulting in the formation of homogeneous residential zones and, in numerous instances, a lifestyle that is more reliant on automobiles (see Plenk (2022)). Consequently, while Germany’s zoning regulations facilitate the creation of integrated, accessible communities, the United States’ model encourages the development of distinct, sprawling residential and commercial zones. This has significant implications for housing supply flexibility and the resulting differences in urban structure and mobility, particularly with regard to commuting patterns. While the average German lives closer to their place of work, a larger proportion of Americans are required to commute longer distances, potentially providing an incentive for relocation during the pandemic and the adoption of remote working practices.^[15] Furthermore, there is no dramatic change in population levels in either the CBD or the suburbs in Germany. We analyzed the population information offered by RWI and Microm (2023), and observed minimal population changes compared to the pre-pandemic levels in 2019. We also used information on migration across city boundaries provided by Federal Statistical Office (2023). The data shows a very stable pattern of movement into or out of German cities for the years 2017–2021. While this data considers movement across cities rather than within cities, which is our focus here, it provides suggestive evidence that residents in Germany do not tend to move frequently and probably less so in uncertain times.

The disparate structures of the U.S. and German housing markets provide a rationale for the divergent donut effect dynamics. In the United States, a robust homeownership culture, market-driven pricing, and single-use zoning contribute to a flexible housing market with greater suburban expansion and easier relocation. This flexibility has resulted in a notable increase in suburban property prices. Conversely, Germany’s rental-oriented culture, protective housing policies, and mixed-use zoning contribute to the stabilization of demand across urban and suburban areas, thereby limiting the potential for significant suburban price increases. Due to the presence of restrictive housing supply constraints, lower relocation rates, and stable population distributions, the German market demonstrates a diminished donut effect, with housing demand remaining relatively consistent across regions.