New methods for estimating employment and business in small or bespoke areas

Planners and policymakers often need job and business counts for areas such as the Central Activities Zone (CAZ), the Northern Isle of Dogs (NIOD), Opportunity Areas, town centres and high streets. Official statistics^[1] for employment (from the Business Register and Employment Survey) are available down to the Lower Layer Super Output Area (LSOA) level and business statistics (from ONS Business Counts) to the Middle Layer Super Output Area (MSOA) level, and bespoke boundaries rarely align with either.

GLA Economics has previously apportioned counts by land share: the share of an LSOA's or MSOA's land area falling inside the target area. This assumes activity is distributed evenly across all types of land. Thus, it over-allocates to areas with large residential or open space and under-allocates to dense office clusters. This supplement summarises a recent GLA Economics methodology paper on refining the employment and business estimation in small or bespoke geographies.

Allocation using land-use data

This paper addresses these limitations by replacing the total physical land layer with commercial land layers that contain building information. We use Ordnance Survey (OS) National Geographic Database (NGD) building data which for each building contains its footprint area, its use class (i.e. commercial or residential) and the height (number of floors). A building counts as commercial if it has at least one commercial address, which includes mixed-use buildings.

Depending on whether we are counting the number of businesses in an area, or the amount of employment, we use either the footprint area or the floorspace area, as follows:

Businesses: Commercial Land Share. The MSOA's business count is allocated by the share of the MSOA's commercial land that falls in the target area.
Employment: Commercial Floorspace Share. The LSOA's employment estimate is allocated by the share of the LSOA's commercial floorspace (footprint × number of floors) in the target area. Floorspace is used because jobs scale with floor area, whereas a business counts as one unit regardless of size. For mixed commercial/residential buildings we assume only one commercial floor.

How well does it work?

We tested each method by apportioning known totals down to areas where observed counts exist, measuring the share of activity placed in the wrong area (lower is better).

Allocating London totals to boroughs, the old method misplaces 47% of employment and 34% of businesses, pushing activity outwards: Outer London boroughs received 132% more employees and 68% more businesses than observed. For employees, commercial floorspace further improves on the commercial land method, reducing the misallocation from 28% to 13%. For businesses, the two refined methods perform similarly, at 18% and 16% respectively.

Within MSOAs, the employee misallocation falls from 29% to 19% when switching from land share to commercial land share. Although the floorspace method does not substantially improve performance, floorspace remains the preferred benchmark because it can be updated when new development comes forward and it is a well-established measure in planning.

The old method undercounted dense Inner London areas. Using the new methods, the CAZ has 3% more jobs and 4% more businesses, and NIOD has 8% more jobs and businesses.

Advantages and limitations

The main advantage is increasing accuracy without (much) added complexity. The new methods keep the simple, transparent mechanics of land share apportionment and use the same official counts as inputs (BRES and Business Counts), but better reflect where activity occurs, reducing misallocation.

Residual error remains. Even the best methods misallocate around 13% of employees and 16% of businesses at borough level, and around 19% of employees across LSOAs within an MSOA. Estimates for very small areas should be treated with caution.

The new methods still rely on distributional assumptions: businesses are spread uniformly across commercial land and jobs uniformly across commercial floorspace, whereas real densities vary by sector and building type. Home-based work and work in non-commercial land is not captured, although this affects only around 0.1% of London’s employees (i.e. excluding the self-employed). Finally, the commercial map is a May 2026 snapshot that does not reflect temporal changes, and floorspace measures capacity rather than occupancy, so vacancies are not accounted for.

A summary of old and new methods is in Table A1.

Allocation using land-use data

How well does it work?

Advantages and limitations

Recommendation