1. Introduction: This is an empirical, quantitative UK study using Rightmove’s online housing search data. Rightmove.co.uk is the UK’s largest online real estate portal and property website, and particularly in England and Wales, holds more than 90% of properties that enter the market. The data analysed is generated by the users of the website who can search for housing through the Draw-A-Search, resulting in over 4.5million visits per day. Users are able to change attributes such as house price range, number of bedrooms, or property type, as well as applying filters such as the property requiring a garden, garage or being a new build property, for example. The drawn searches require the users to match their mental maps with respect to Google map data in the Rightmove website. The aim of the project is to assess the extent to which online housing search data can help us identify, spatially, housing market and sub-market areas through the use of user-generated search data. Utilising Rightmove’s search data will result in an analysis on the gap in knowledge of the geography of housing search in relation to the geography of choice and demand. This paper focuses on Greater London in order to increase understanding of the user-drawn housing market in the UK.

2. Conceptual Framework of the Study: There is a gap in research in how people search for housing online, and the supply and demand statistics related to this search. It assumes that those people who are actively searching for a house all want the same attributes of buying and selling houses (number of bedrooms, price, moving in/out date) as everyone else in society which creates a market equilibrium- but this is not the case. There is already substantial previous research on the housing market, yet the majority of the research is prior to 1990. This is consequently outdated as since then the prominence of the online housing market has emerged. Publicly available Google trends data and online search data have thus advanced this method to that of prediction locations (Beracha and Wintoki, 2013; Choi and Varian, 2012; Wu and Brynjolfsson, 2009). This has proven to be of little avail as the majority of research draws to nowcasting instead of predicting the future, which is where the gap in literature and methods lies. Research has tended to focus on where people have moved to, as opposed to where a searcher might want to move to (Rae, 2015) so this project will delve deeper into this latent versus revealed demand argument.

3. Study Area and Data: The project focuses on Greater London in the UK. Greater London, for this study, will be treated as a separate case, given its uniqueness within the English housing market and its significantly different characteristics with respect to price and geography, which Gray (2017) also defended. The aim of the project is to assess the extent to which online housing search data can help us identify, spatially, housing market and sub-market areas through the use of user-generated search data. The project takes a quantitative approach, using deductive reasoning, empirical evidence and hypothesis testing. The project primarily uses secondary data provided by Rightmove plc. The data is generated by users of the website searching for properties. The data analysed focuses on the Draw-A-Search Tool. The data used is from the month of March from 2012-15 for the drawn searches as March is when people typically choose to buy a house so there are more data to analyse. The formula used for the analysis () is called the Polsby-Popper Ratio which gives a result between [0,1]. A score of 1 means that it is a complete circle and has maximal compactness, whereas a score of 0 means that there is a complete lack of compactness. This analysis conducts how irregular a shape is drawn within the Rightmove Draw-A-Search Tool. Some examples are shown in Figure 1, where it can be seen that from left to right the last four shapes diverge from a perfect circle. It has been assumed that the lower the score, the more knowledge a user has about a location. This will be explored in the analysis.

4. Results: Figure 2 shows that the majority of drawn searches within the Greater London boundary from March 2015 fall within the 0.7-0.8 ratio, meaning that most users are drawing regular shapes such as rectangles, than opposed to irregular shapes that look like paint splatters. Figure 3 shows a violin plot for 2012-2015 compactness scores in London, and it shows that there is little change in the way users are drawing shapes. It can then be assumed that the majority of users are first time buyers within London, as opposed to moving within London. A very small amount of users (~1,000 polygons) have a ratio of <0.1 which means they have based their search around the London underground stations, green parks or specific streets. This can conclude that these users have good prior knowledge of the desired property location. The data also reveals where the polygons overlap, as shown in Figure 5, and this shows where the most drawn demand is and this has been split into detached, semi-detached and flat property types. It shows that within London, the most drawn property type is a flat and this covers almost all of the Greater London boundary. A smaller scale can also be applied, as shown in Figure 4. This shows a user drawn shape along the River Thames so they wish to live in a property with a river view. Similar results have been found along bus/tram/underground/train routes, and also ommiting specfic areas such as whole housing estates.

5. Conclusion: This new source of data therefore also allows us to explore from a technical point of view how the technology of housing market search might be helping shape or influencing peoples search behaviour. One of the ways that this can be achieved is by looking at the way in which people draw shapes, and how intricate they are compared to how big they are. To interpret the intricacy of the shapes drawn as a measure of how engaged or specific people are in their search some geometry is needed. The Polsby-Popper score gives some indication of the intricateness of shapes which is explained in the results. It it is also interesting to consider whether technology is hindering or helping the online housing search, and whether their knowledge of mental maps is suitable to finding their dream property. The majority of people are searching in broad, larger areas, as opposed to smaller intricate drawn shapes that limit the number of properties returned. It is then estimated that most people are moving to new locations that they are not familiar with, as opposed to staying in the same location. The search versus actual sales has been correlated on a six-month lag and has a strong positive correlation, so it can be concluded that the drawn search demand can accurately predict the future sold property outcomes.