In a sector which is highly heterogeneous, opaque and unorganised, developing granularised turnkey solution to represent the geography holistically in the form of an index was a huge challenge.
The entire exercise required collecting primary market data through field visits and surveys, factoring in circle or ready reckoner rate, stamp duty, alongside prices quoted by independent brokers and agents.
Before finalising algorithm to create the index, it was important to assign importance to each data set.
The two exhaustive tasks that we were expected to crack in a short span of time were:
Creating end-to-end automated enterprise solution
The entire process that goes behind data collation, data sorting and formulation of indices has been streamlined and automated to minimise needs of
manual intervention and possibility of error. Liases Foras has created a workflow for the purpose of data collection and extraction from multiple sources and data cleaning to maintain highest level of accuracy during the preliminary analysis. Before converging data from varied sources and formats into a singular format, the enterprise solution created for
NHB automatically detects anomalies and errors at the time of revalidation.
Finding contemporary methods to compute indices
Coming up with hedonic regression to determine importance of variety of factors and characteristics governing housing market was a tough nut to crack. While modified Laspeyres method was being used earlier, we decided to stick to weighted average prices before using Laspeyres formula to calculate Housing Price Index at Market Prices (under construction properties) and Housing Price Index at Assessment Prices (property value maintained by housing finance companies and institutions).
To arrive at right estimates, we changed the entire approach and decided to compute index as per weights assigned to different products. In previous versions of NHB RESIDEX, transactional weightages were assigned to boundaries/locations within a city. The practise limited the flexibility to assess prices at various regional boundaries without distorting the city level prices. Prices at city level should remain constant irrespective of the number of boundaries.
For example, Greater Mumbai has 24 Municipal Wards, 97 Census Wards/ Sections and 104 Pin Codes. Housing Price Indices of a city, when computed using weights of Municipal Wards would be different from Housing Price Indices computed based on weights of pin code.
As city prices cannot change with change in boundaries, we decided to do away with weightages of regional boundaries. Price should be the weighted average price irrespective of selection of boundary which may be a region, city, pin code, ward, a micro market or colony.
Hence weightages were based on number of transactions in the base year as per different product categories to compute revised RESIDEX. In addition to transactional weightages, factors using housing / population stock weights have been also applied at zonal and product levels.
The formula for computation is as: wherein, P0i = Price of i th product in base period Q0i = Quantity of Unsold stock/ Number of transactions of i th product in the base period P1i = Price of i th product in the current period n = Number of product types.
To make the entire process contemporary and attuned to popular credit schemes, product category classification is based on carpet area size under Credit Linked Subsidy Scheme (CLSS) guidelines. The three product categories are (a) units measuring less than or equal to 60 sqmt (b) units measuring more than 60 sqmt but less than or equal to 110 sqmt and (c) units measuring more than 110 sqmt. The main reason behind choosing this classification is to map the impact of affordable housing schemes that are being run by the government.