How to beat numbers and not let numbers
How to beat numbers and not let numbers beat you Professor Charles Pattie
Why do we need numbers in social science? • Why do we distrust stats – and should we? • Samples • Why do we need numbers in social science? • Context: how common or unusual are things we study? • Trend: are they increasing or decreasing? • Correlation: what are they related to, and what might influence them?
So why do we distrust stats? • Lies, damn lies and statistics? • Reliability: look at the question and the context • North Korean election 2014: 100% vote for Kim Jong-Un! • Common sense rules: distrust clearly biased sources!
So why do we distrust stats? • Lies, damn lies and statistics? • Accuracy: how close can we get? • Sampling: a statistic is only as good as the sample it is based on • An analogy from cooking: tasting a small sample tells you what the whole dish is like
Sampling in practice • The central limit theorem: why random samples ‘work’:
Sampling in practice • Rules of thumb – when to trust a statistic. • Is the sample (more or less) random? (BIASED samples are bad news) • Is the sample large enough? (Think Goldilocks – not too large, not too small) • Does the question make real sense? (NB NOTHING to do with stats, EVERYTHING to do with common sense!) • If there was an election tomorrow, which party would you vote for? • On which day of the year are people most likely to feel unhappy? • Do you think the local authority’s urban development policies are successful or not?
Context: Is crime out of control? • Are we in the grip of a crime explosion? Discuss!
Context: Is crime out of control? • The National Crime Survey suggests not:
Context: Is crime out of control? • But do we believe it? Some views from BBC News ‘comment’ section:
Context: Is crime out of control • Why are people so sceptical? Discuss! • Anecdote versus evidence?
Trends: A world of numbers? • Seeing the wood for the trees: the value of careful presentation • e. g. the Gapminder project: http: //www. gapminder. org/
Correlations: A different example…
Correlations: UKIP and the 2014 Euro-election • Why did Nigel Farage launch UKIP’s 2014 European Parliament election campaign in Sheffield? • Should David Cameron (and Ed Miliband Nick Clegg) be worried about the Euro-election? • Can numbers help answer these questions?
Why did UKIP launch their campaign here? • What do YOU think? Hunches? Ideas? • Might it have something to do with who votes UKIP? • Evidence from British Election Study Continuous Monitoring Survey 2004 -2013 (reported in Rob Ford and Matthew Goodwin, 2014, Revolt on the Right, Routledge).
Why did UKIP launch their campaign here? Class and party support 50 45 40 % support 35 30 25 20 15 10 5 0 UKIP Conservative Professionals Routine non-manual Labour Working class Lib Dem
Why did UKIP launch their campaign here? Education and party support 60 % support 50 40 30 20 10 0 UKIP Conservative Left school aged 16 or under Labour Left school aged 17 or 18 Lib Dem Left school aged 19+
Why did UKIP launch their campaign here? Age and party support 60 % support 50 40 30 20 10 0 UKIP Conservative Under 35 Labour 25 -54 55+ Lib Dem
Why did UKIP launch their campaign here? Age and party support 100 90 80 % support 70 60 50 40 30 20 10 0 Govt doesn't treat people like me Big gap between what I expect Immigration most important issue fairly (% agree) and what I get (% agree) UKIP Eurosceptics Non-UKIP Eurosceptics Europhiles
Why did UKIP launch their campaign here? • So… what do you think?
Correlation: lessons of 2014 Euro-election? • Should David Cameron (etc. ) be worried? • Maybe! You. Gov poll, 22 -24 April 2014:
Correlation: lessons of 2014 Euro-election? Vote intention 40 Euro-election 35 Westminster election % support 30 25 20 15 10 5 0 UKIP Labour Conservative Lib Dem
Correlation: lessons of 2014 Euro-election? Vote intention 40 Euro-election 35 Westminster election % support 30 25 20 15 10 5 0 UKIP Labour Conservative Lib Dem
Why the switch? • Second order election theory: Euro-election a risk-free chance to vent frustrations with government
But is the 2 nd-order effect getting stronger (or are governments just getting less popular)? 55 55 45 40 35 30 25 20 15 1979 Westminster 50 % support for incumbent 50 Europe 45 40 35 30 25 20 1989 1999 2009 15 1979 1989 1999 2009
Moving on…. • We’ve only scratched the surface of using numbers in social science research. • Where next? • Getting good data • Collect it yourself? • Data archives and official sources e. g. The UK Data Archive (http: //www. dataarchive. ac. uk/) • More advanced techniques and software • Good introductory guides include: Ian Diamond and Julie Jeffries, 2001, Beginning Statistics: A Guide for Social Scientists, Sage. • If you learn one more advanced method, learn REGRESSION! • SPSS available ‘free’ to university students • Help at MASH
Moving on…. • And with that – happy analysing!
- Slides: 31