February 26, 2024

5 min read

Parcl Labs Price Feed White Paper Paris V2

Executive Summary

  • In July of 2023, we released the first version of the Parcl Labs Price Feed (PLPF) for Paris, an indicator that tracks daily price changes in residential real estate in the city of Paris as measured by price per square meter in Euros.
  • After a few months of evaluation, we are releasing an improved version that better reflects the residential real estate in Paris. We achieve this with an enhanced weighting schema where official historical sales and more timely data have a more gradual approach on how the weights are combined. Further, we modify our filtering parameters to obtain a more representative sample of real estate transactions by increasing the sample space of our approach.
  • We improved our data pipeline to ingest, clean, and transform millions of individual data points across multiple sources, leveraging spatial data science to generate precise price estimates for Paris. We have used this approach to build price feeds for multiple cities in the USA and London.
  • We ran thousands of experiments to further enhance the accuracy and relevance of our price feed and to determine the optimal parameters for our smoothing time series techniques. These techniques are a continuation of the gold standard defined in our Parcl Labs Price Feed White Paper.
  • The PLPF for Paris empowers users to make better-informed real estate and investment decisions. Contact our team today if you want to access Paris and other global market price feeds.

Figure 1. Parcl Labs Paris Pricefeed V2

Introduction

The French real estate market encounters challenges similar to those in the United States, characterized by fragmented and opaque data. Timely information, such as property listings, is often incomplete and scattered. Additionally, crucial details about properties, like their precise locations, are not easily accessible. Official data sources such as the Demandes de valeurs foncières also fall short in providing timely and comprehensive market insights, with a lag of more than 6 months from closing to publication along with incomplete property details. This information gap leaves the general public and customers dependent on piecemeal and insufficient data for making informed decisions.

Even when real estate information is available, such as with the Paris SQM Index developed by Compass, it suffers from a two-week lag and is derived from a privately owned database that requires payment for access with unclear guidelines on how to qualify to obtain the data . Further, their data cleaning and index construction considers observations that can potentially bias the price per square meter in Paris as they seldom label outliers as such.

Today, we are excited to announce an updated version of our Parcl Labs Price Feed for the City of Paris. This release builds upon our previous Paris Price Feed, offering optimizations to better reflect the current market conditions of residential real estate in the French capital.

Data

The segmentation and incompleteness of data pose significant challenges in creating reliable price feeds in France. To address these challenges, we rely on historical data published by the Government of France, along with more timely information from real estate property listings. We focus on residential properties, including condos, apartments, single-family homes, and others, allowing us to combine historical trends with more up-to-date data.

Similar to the Parcl Labs Price Feed, we employ a rigorous process to create a top-notch data warehouse for the development of our Price Feeds. This process involves:

  • We undergo a process of cleaning and deduplicating millions of observations to weed out duplicate data entries and ensure data integrity. This allows us to use more than 6 million data points to calculate our price feed.
  • Once the data is cleaned, standardized, and processed, we leverage state-of-the-art geographic information systems technology to assign properties to multiple types of markets. This enables us to select specific information for desired markets and facilitates tailored analysis at various geographical levels, such as arrondissements.

Figure 2. Map displaying data points used to calculate the Parcl Labs Paris Price Feed.

In Version 2 of the price feed, we introduced additional filters for property selection and enhanced our deduplication process to focus exclusively on unique price points, thereby offering a more accurate representation of the market.

Methodology

We follow the same general principles outlined in our original price feed methodology where we describe our approach of building time series data with the millions of data points available in our warehouse. The main difference between this version of the pricefeed is the use of different shorter back-propagation windows based on the volatility of transactions. This simply looks at how many transactions in a given period of time are available in each market before deciding how far back in time we are going to look to create a sample space. The window selected changes by market, to reflect local idiosyncrasies and capture the relevant data volume. Through the simulation of thousands of scenarios, we have fine-tuned our model's parameters.

In version 2 we also fine tuned the configuration of percentiles for our sample space. In Version 1 of our price feed, we utilized observations falling between the 25th and 75th percentiles of the price distribution. After running thousands of simulations with varying percentiles, we concluded that the new version will incorporate observations between the 10th and 80th percentiles. This approach allows us to better reflect the distinct dualities within the Paris real estate market, particularly in the more affordable segment.

Figure 3. Distribution of Sales in Paris in 2022 and first half of 2023 as Measured in Price per Square Meters in Euros

After conducting initial estimates using both timely and historical data, we apply a weighted average that incorporates decaying geometric weights for periods with overlap between historical sales and more recent data. This approach enables us to achieve a sample that more accurately reflects the real estate market, blending short-term insights with longer-term trends. The final stage involves a smoothing process through a moving average, specifically tailored to each market, to mitigate price fluctuation anomalies. In Version 2 of the price feed, we have adjusted the parameters of this final smoothing process, opting for a shorter window for our moving average.

To ensure data consistency and reliability, we perform market-specific testing for each market covered by our API before releasing a data update. This testing addresses potential anomalies across different data sources, local market peculiarities affecting volume and price volatility, and geographic factors that might influence the volatility of our series. The result is a time series that rigorously evaluates any abrupt price changes.

As an additional quality control measure, we also calculated the Pearson correlation coefficient for the Parcl Labs Price Feed series and the Paris SQM from August 4th 2017 to February 23th 2024 and found a correlation of 0.96. This high correlation further validates the robustness and accessibility of our daily data.

Figure 4. Correlation Between Parcl Labs Price Feed Paris V2 and Paris SQM Index for Paris 2017-2024

Conclusion

The fragmentation and lack of transparency in real estate data in the Paris market have prevented users and consumers from making informed decisions about residential real estate. Incomplete and siloed information has long been the norm.

Today we are updating the only available daily price indicator of residential real estate available for Paris. Our new price feed improves the timeliness of our data, and better captures the bifurcated nature of the Parisian market while maintaining the gold standard of our first price feed.

With our user-friendly API, the Parcl Labs Price Feed for Paris puts the power of informed decision-making directly into the hands of our users. Get ready to make confident real estate choices like never before.

Subscribe to our newsletter
Read about our privacy policy.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Share this post