Data on Repeated Offerings in the German Housing Market Based on RWI-GEO-RED

1 Introduction

Expenditure on housing costs dominates the expenditure structure of most households, both in terms of monthly rent payments and, in many cases, the cost of property financing. Turning to assets, for most owner-occupied households, housing is the largest household asset. Individual households (as current or potential future owners) tend to pay close attention to changes in house prices that affect wealth or affordability respectively (OECD et al. 2013). Because of its importance in terms of costs and assets, real estate price research is therefore a well-established topic in many economic research papers.

Here, residential property price indices (RPPIs) play an essential role, serving economists and policy makers as an indicator of financial soundness and a variety of other economic activities. By its very nature, housing is a highly heterogeneous good without a single observable market price (Leishman and Watkins 2002). Thus, attempts to transform information on properties with a variety of price-determining characteristics into a single representative measure face a number of obstacles. A perennial problem in the construction of RPPIs is differences in the quality of the characteristics (Englund, Quigley, and Redfearn 1998; Greenlees 1982; Kirby-McGregor and Martin 2019). These differences are of two types: improvements and depreciation. In times of an increasing need for energy renovations to impede climate change, the former has only gained in relevance. More generally, price movements due to quality changes needs to be disentangled from purely time-related movements in order to allow unbiased comparisons.

To achieve this, the repeated-sales methodology exploits multiple sales of the same unit over time and calculates the price difference between sales as the constant quality price movement of the unit. This approach attempts to control for time-invariant unobserved/unobservable differences in quality between properties, a problem that typically arises when using hedonic models (Englund, Quigley, and Redfearn 1998; Nagaraja, Brown, and Wachter 2014), e.g. curb appeal or the quality of landscaping (Bajari et al. 2012). First proposed by Bailey, Muth, and Nourse (1963, geometric repeated sales-index, GRS) and later adapted by Shiller (1991, arithmetic repeated-sales index, ARS), the methodology is widely used in past and current literature and is even referred to as the “the gold standard in house price construction” (Ahlfeldt, Heblich, and Seidel 2023, p. 1). In the context of Germany, however, no data are available to construct such indices and thus no opportunity exists for research to benefit from the potential advantages of this methodology. In this paper, we propose an algorithm to classify repeat-sales and repeat-rent in the German real estate market, based on information on real estate listings obtained from Germany’s largest real estate listing website, ImmobilienScout24.de, thus filling the gap in the literature.

The article is structured as follows: Section 2 describes the data basis and the algorithm, while Section 3 shows descriptive statistics of the resulting data and uses them to construct indices. Thereafter, Section 4 outlines some advantages of the data and suggests potential research avenues. Section 5 concludes with information on how to access the data.

2 Repeated Sales Classification

3 Descriptive Statistics and Application

As mentioned above, a popular application of repeated sales data is the construction of price indices. In this section we compare the hedonic price index, the current standard for Germany, with the repeated sales indices (Bailey, Muth, and Nourse 1963; Shiller 1991), whose construction is enabled by the new data. To ensure comparability, the data for both indices are prepared in the same way, following Thiel (2024b). This RWI-GEO-REDX dataset on housing price index is described in Klick and Schaffner (2021). In addition, the quarterly index values are based on the first quarter of 2010 to allow for comparison of values over different magnitudes. Therefore, each change is expressed in percentage points relative to the corresponding base year values.

Figure 1 shows the different price indices for apartment rents in Germany between 2010 and 2023. The hedonic index (blue) shows the strongest upward trend, indicating a rent increase of about 72 percentage points. The repeated sales indices, ARS (red) and GRS (green), paint a more conservative picture of total rent growth over the time horizon, suggesting growth of 44 and 46 percentage points, respectively. Nevertheless, the overall trend is very similar. All of the price indices show a monotonic increasing trend, where the growth is slowing down, but still remains positive around 2015 and 2020.

Figure 1:

Price indices for Germany – apartment rent. Notes: Illustration of quarterly price indices based in the first quarter of 2010. Source: Authors’ graph.

This pattern is confirmed when zooming into the exemplary cities of Berlin, Frankfurt, and Munich in Figure 2. The cities are colored as indicated in the figure legend, where the dotted line shows the hedonic and the solid line the GRS value. For each city, the hedonic measure indicates a larger growth, while the GRS estimate the price increase more conservatively.

Figure 2:

Price indices for selected cities – apartment rent. Notes: Illustration of quarterly price indices based in the first quarter of 2010. Source: Authors’ graph.

In Figure 3, we repeat the illustration of the three indices for apartment sales instead of apartment rents. Again, the pattern is repeated, with monotonic increases in all three measures. While the hedonic index still shows the largest growth rates, the difference is considerably smaller in magnitude. Notably, the growth rates for apartment sales do not seem to have experienced a periodic slowdown, but rather an ever-increasing trend.

Figure 3:

Price indices for Germany – apartment sales. Notes: Illustration of quarterly price indices based in the first quarter of 2010. Source: Authors’ graph.

4 Advantages and Research Avenues

The extended data offer two main advantages to researchers. First, the construction of price indices similar to those described above for different regional units. Second, the emergent panel structure resulting from the clustering of the longitudinal data. This allows the study of common and individual dynamics of units, such as listing updates as well as mispricing (Knight 2002; Knight, Sirmans, and Turnbull 1994; Lyons 2015). Furthermore, it allows for unit-level fixed effects to be included in models. This is much more precise than other common alternatives, such as grid-level fixed effects, and can therefore control for a much more detailed range of unobserved/unobservable heterogeneity. For example, the fact that an apartment faces north and therefore has little natural light cannot be observed by the researcher, but it can reduce the value of the apartment.

5 Data Availability

The variables introduced by the algorithm will be added to future releases of the RWI-GEO-RED data, while the code can be found at [GitHub].^[3] The data are available to researchers for non-commercial use. There are two versions of the RWI-GEO-RED data. First, the Scientific Use Files (SUF) cover all information except the exact geo-coordinates. Second, the full datasets are available in the Data Secure Room of the FDZ Ruhr in Essen (on-site access).^[4] The data can be obtained as a Stata datasets (.dta), .csv files, andparquet files. Data access to both versions requires a signed data use agreement. Both versions are restricted to non-commercial research and only researchers of scientific institutions are eligible to apply for data access. The SUF may be used at the workplace of the users. Data access is provided by the Research Data Centre Ruhr at the RWI – Leibniz-Institute for Economic Research (FDZ Ruhr). The data can be accessed at the data access website of FDZ Ruhr.^[5] The application form includes a brief description and title of the project, information on the applying department, expected duration of data usage as well as further participants in the project. Data users shall cite the datasets properly with the respective DOIs.