Education, Immigration and Brexit

Education, Immigration and Brexit

Does diversity of students in school improve everyone's achievement ?

Overview

What was made: 

Four maps displaying information on:

  •       Proportion of voters choosing 'Leave' in EU ref. 2016
  •       Proportion of students getting 5 Cs or better on GCSE exams for 2015-16
  •       Proportion of UK residents who are born outside UK
  •       Proportion of students who are not 'white British' 

   Two graphs showing correlation:

  •        Leave voters and immigration
  •        School diversity and GCSE results

 *Created for England and London, separately.         

    Background and objectives: 

    At a regional level, intensity of immigration correlates positively with educational achievements and desire to remain in the EU. Areas which voted for Brexit seem to correlate with the relative absence of immigrants and worse school results. Is immigration a cause or are there other socio-economic factors that explain the pattern better? ​We started by discussing and reviewing research on a relationship between immigration, school performance and EU referendum.  The objective was to visually document this complex relationship in a form of maps.  We considered publically available data in relation to geography, focusing separately on England and London. Economics, class and cultural differences - all play a role in influencing these data.  Looking at different parts of UK, as we have done, shows how easy it is to come to oversimplified conclusions: the positive correlation between immigration and academic achievement which is apparent in England's data disappeared if we focus on London, using a similar level of data aggregation.  Controling for income, educational level of parents, class sizes etc would make for a more robust statistical analysis.  Data is not a replacement for in-depth research.  

    Outcome: 

    There appears a geographical negative correlation between EU voters and proportion of people born outside UK. There seems to be a positive geographical correlation between proportion of people born outside UK and GCSE performance.  Correlation is not an indication causality, and the first claim in particular has been disputed in previous research [1,2]. In particular, several studies claim that it is not the proportion of immigrants in the area that has consequences for political attitudes but the relative change in their numbers [2].  Moreover, a higher level of immigrants in a community often corresponds to larger metropolitan  areas with younger  population  and a range of more hopeful socio-economic conditions.  There is an argument that 'leave' voters are more likely to be found in areas experiencing economic decline - the areas into which immigrants are less likely to move into.  The same economic arguments can explain the correlations between ethnic diversity and school achievements. Cultural differences as well as class differences have been used to explain why so many immigrant's children outperform working class whites. Many immigrant families, particularly from Asia have lower levels of divorce, value education and raise their children with greater emphasis on discipline and scholarliness [3].  The negative correlation between proportion of immigrants and desire on behalf of white British people for fewer of them has been explained by various sociologists in terms of mourning for a lost sense of relevance and a superior position in social hierarchy:  ahead of women, non-whites, and non-Christians [1].  In UK, those who have voted to 'leave' may have perceived themselves as losers in the new economy that is shifting in favour of metropolitan areas. The 'leave' voters tend to be older, nostalgic for the days of the Empire and feeling betrayed by the liberal 'elites' those values (such as anti-racism, anti-sexism, anti-homophobia, etc) they don't always share.  References [1] A. Hochschilds 'Strangers in their own land: fear and mourning on the American right', The New Press, 2016.[2] A JRF report reviewing similar studies https://www.jrf.org.uk/report/brexit-vote-explained-poverty-low-skills-and-lack-opportunities[3] Parliamentary Report on Underachievement of White British Children  https://www.parliament.uk/business/committees/committees-a-z/commons-select/education-committee/news/white-working-class-report/

    How it was made

    Find data sources (accessed 03-12-2016)
    Feeling during this step: 
    5
    No
    Clean data using R studio
    1. Prep GCSE and Ethnicity data for use with immigration and referendum results using RStudio.
    • #####################################################
    • ## Cleaning data set on GCSE results for UK pupils ##
    • #####################################################
    •  
    • ## Load libraries
    •  
    • library(dplyr)
    • install.packages("xlsx")
    • library(xlsx)
    •  
    • ## Read in csv file of GCSE results for 2015-2016 and subset the percentage of 
    • ## students earning 5 or more A* to C results in English and Maths GCSEs.
    •  
    • GCSE_results_20152016 <- read.csv("2015-2016-england_ks4provisional.csv")
    • GCSE_1516 <- select(GCSE_results_20152016, c(2, 15))
    • names(GCSE_1516) <- c("Old.LA.Code", "5.or.more.A*-C.in.both.English.and.Maths.GCSEs")
    •  
    • ## Unzip csv files with definitions for GCSE results file and read in LA and 
    • ## Region codes.
    •  
    • unzip("Performancetables_174238.zip")
    • LA_Region_Codes <- read.csv("./2015-2016/la_and_region_codes_meta.csv")
    • arrange(LA_Region_Codes, LEA)
    • names(LA_Region_Codes)[1] <- "Old.LA.Code"
    •  
    • ## Merge GCSE results with LA and Region Codes.
    •  
    • GCSE_codes_1516 <- full_join(GCSE_1516, LA_Region_Codes, by = "Old.LA.Code")
    •  
    • ## Download and read in file mapping old LA codes to new LA codes. Merge with 
    • ## GCSE results.
    •  
    • regioncodes <- read.xlsx("nlac_2011.xls", 1)
    • GCSEs <- full_join(GCSEs_results, regioncodes, by = "Old.LA.Code")
    •  
    • ## Prep for integration with immigration and Brexit data sets in Tableau.
    •  
    • names(GCSEs)[6] <- "Area Code"
    • write.table(GCSEs, file = "GCSEs.csv")
    •  
    • ###################################################
    • ## Cleaning data set on ethnicities of UK pupils ##
    • ###################################################
    •  
    • ## Load libraries
    •  
    • library(dplyr)
    • install.packages("xlsx")
    • library(xlsx)
    •  
    • ## Download and read in UK govt data set on ethnicities of students by Local Authority 
    • ## (LA) code for 2013.
    •  
    •                 207733/Local_authority_and_regional_tables_-_SFR_21_2013.xls"
    • download.file(dataURL, destfile = "LAandRegtab.xls", mode = "wb")
    • ethnicity <- read.xlsx("LAandRegtab.xls", 14)
    •  
    • ## Clean data set: select only columns on LA Codes, number White British pupils, and 
    • ## number of All Pupils. 
    •  
    • ethnicity <- slice(ethnicity, 7:194)
    • ethnicity <- slice(ethnicity, 1:175)
    • ethnicity <- select(ethnicity, c(2, 6, 30))
    • colnames(ethnicity) <- c("Area Code", "White British", "All Pupils")
    • ethnicity <- slice(ethnicity, 3:175)
    • ethnicity <- slice(ethnicity, 4:173)
    •  
    • ## Remove rows without data on specific LA.
    •  
    • ethnicity <- ethnicity[!is.na(ethnicity$`Area Code`), ]
    •  
    • ## Change class on numbers of pupils from factor to numeric.
    •  
    • ethnicity$`White British` <- as.numeric(as.character(ethnicity$`White British`))
    • ethnicity$`All Pupils` <- as.numeric(as.character(ethnicity$`All Pupils`))
    Feeling during this step: 
    3
    No
    Importing Data into Tableau and visualising it
    • Import into Tableau and experiment with displaying data on maps and scatterplots.
    •  Challenge: Creating a filled-in map on tableau rather than points.
    • Figuring out what goes ont he dashbaord 
      • Maps
      • Correlation graphs
      • * This step is more a trail and error method to get what you want, there are no designated steps to go there, just play!
    Feeling during this step: 
    0
    No