Update: Why Has London Lost so Many Pubs?

Change in percentage Muslim is only a consistent predictor of change in the number of pubs when excluding the borough of Redbridge; change in percentage foreign-born is a strong predictor of change in the number of pubs

[Update: further analyses available here do not support the hypothesis.]

Further to my previous post, here I present some multivariate models of percentage change in the number of pubs across London boroughs. I obtained data on the following variables: total population, percentage Muslim, percentage aged 0–15, percentage aged 65+, percentage foreign-born, percentage unemployed, percentage with no qualifications, percentage in social housing, and median house price. Data for 2001 were taken from the ONS. Data for 2015 were taken from the London Datastore.

Percentage unemployment, percentage with no qualifications and percentage in social housing were combined into an index of deprivation by extracting the first principal component from a PCA, separately for the 2001 and 2015 data. This component explained 69% of the variance in the 2001 data and explained 61% of the variance in the 2015 data.

The table below shows the correlations between percentage change in the number of pubs and changes in the other variables. Change in percentage Muslim, change in percentage aged 65+, change in percentage foreign-born, and change in median house price all have significant correlations with percentage change in the number of pubs. Notice, however, that the correlation with change in median house price is in the opposite direction from what one might have expected: boroughs where the average house price has risen more seem to have lost fewer pubs.

The table below shows estimates from linear models of percentage change in the number of pubs, using change in percentage Muslim as the predictor of interest. While it has a negative sign in all five models, it is only a significant predictor in models 3 and 4. In model 5, which includes all 7 covariates, only change in percentage foreign-born and change in deprivation index are significant predictors.

The table below repeats the above analysis excluding the borough of Redbridge, which was shown to be something of an outlier in my previous post. Here change in percentage Muslim is a significant predictor in all five models. Change in percentage foreign-born also reaches significance in models 2 and 5.

The final table shows estimates from linear models of percentage change in the number of pubs, using change in percentage foreign-born as the predictor of interest. It is a significant negative predictor in all five models.

None of the other variables was a consistent predictor of percentage change in the number of pubs.