Five questions about Dubai’s 2026 market

A fixed-effects regression of log(price per sq ft) on month, with one fixed effect per area × type × bedroom group, controlling for size and off-plan status. Each month’s coefficient is the like-for-like change compared with Jan.

Implemented with the within transformation (subtract each group’s mean, then least squares), which gives the same answer as one dummy variable per group.

Headline: the off-plan coefficient of the fixed-effects regression above (same area, type and bedrooms; controls for size and month).

By area: within each type + bedroom group with at least 15 off-plan and 15 ready sales, the ratio of median prices per sq ft, averaged across groups. New off-plan buildings also tend to be newer and better equipped, so the premium includes that.

For each area with at least 20 sales of both, the median price per sq ft of each bedroom count divided by the 1-bedroom median, averaged across areas (weighted by number of sales).

The size coefficient (elasticity) of the same fixed-effects regression: log(price per sq ft) on log(size). Apartments: -0.38, villas: -0.414. 1.1 raised to that power gives the change for a 10% bigger home.

5-fold cross-validation grouped by project (GroupKFold), so a project is never in the data its prediction was trained on. The score is the median of actual ÷ predicted price over the project’s sales. Cross-validated median error: 9.9%.

A premium usually reflects things the data doesn’t contain: brand, views, finishing, amenities or payment plans. A discount can mean an older or smaller building, a weaker location inside the area, or unusual deals. This is a starting point for research, not a buy or sell signal.