Anest

Technology

The project is written as a series of R packages, in turn using a large set of R packages from the CRAN site. The results are refreshed monthly from raw data to generate a full set of reports including this website which is written in markdown/blogdown and auto-refreshed monthly. The source data and software utilised are freely available under some form of Free/Open Source licence.

Giles Heywood - project author

After reading Natural Sciences (Physics) at Cambridge I developed signal processing algorithms at a geophysical consultancy in the oilfield seismic and well log analytic sector. An MBA and later MSc in econometrics led to quantitative fund management at city institutions first as analyst, later as fund manager, then to senior positions including head of quant research at Gartmore and head of single-strategy hedge funds at ABN AMRO Asset Management. I developed a taste for R programming as senior analyst at Commerzbank Risk Analytics group, where we delivered portfolio analytics to institutional clients.

The Anest project applies a range of technologies to extract signal from large bodies of data - I continue to study latest quantitative best practice and am currently completing the ARPM marathon with certification, whilst consulting to property PE clients with Bloomberg data analysis.

Seven Dials Fund management principals Brett Robinson - formerly a PM colleague at a London quant boutique - and Mickola Wilson introduced me to the Land Registry Price Paid dataset, and their countless suggestions and observations have been invaluable. Thanks also to Gavriel Merkado at Realyse for my internship. The material here has been presented in earlier versions at Investment Property Forum and Thalesians in London.

Outputs

A full set of outputs is available on request - this is a selective list of some key results.

  1. indexes for 104 postcode areas in England and Wales, plus district and sector-level indices where these have demonstrably higher accuracy in explaining prices, which they mostly do.

  2. revalued prices for all properties sold at least once since 1995

  3. a high-accuracy join of EPC and Land Registry addresses, price and floor area data for price/m2 analysis, aggregated and disaggregated

  4. risk models based on factor analysis and copula at area, district, sector granularity

  5. scenario-based projections and stress tests

  6. case studies of performance attribution e.g. Crossrail, New Build Premium

Contact: inmail on linkedin

Price data from Land Registry Price Paid Data

Coordinates from Ordnance Survey

Postcodes from the Post Office

R is an open source software project

Hosted by Netlify with help from Hugo and styled with Learn/Grav