Presentation is loading. Please wait.

Presentation is loading. Please wait.

Caitlin Milazzo University of Nottingham Jesse Hammond

Similar presentations


Presentation on theme: "Caitlin Milazzo University of Nottingham Jesse Hammond"— Presentation transcript:

1 Crowdsourcing campaigns: A new dataset for studying British parties’ electoral communications
Caitlin Milazzo University of Nottingham Jesse Hammond University of California, Davis

2 We know a lot about what British parties do to attract voters…

3 But we know less about what parties say to them
Providing information one of the key components of a local campaign Parties spent more than £15 million on ‘unsolicited materials’ (40 per cent of all campaign spending) Leaflets the most of common form of election contact (2015, BES) Parties strategic about what they do, so perhaps also about what they say

4 Gathering data on election communications is costly
Shephard (2007) – Scotland, 2001, 2005 Contact with election agents But small no. of communications (2001 N = 70, 2005 N = 51) Fisher (2005) – 223 constituencies in England, Scotland, Wales Volunteers in each constituency collected every election communication delivered 3,459 election communications

5 Using crowdsourced information to lower the costs

6 Building the dataset Python-scraping to gather images, meta-information Some leaflets discarded (local elections, incomplete leaflets) Major, widely-competitive parties CON, LAB, LD, UKIP, Greens, and SNP → All won > 1 million votes Insufficient leaflets from PC Additional (manual) coding 21 policy dimensions (UK Policy Agendas Project) Candidate traits Images Mention of opposing parties/candidates

7 Distribution of election leaflets
Party Count Per cent Conservative 750 22.70 Green 373 11.29 Labour 891 26.97 Liberal Democrats 727 22.00 SNP 96 2.91 Ukip 467 14.13 N (leaflets) 3,304 100.00

8 Distribution of leaflets
429 constituencies (68 per cent) No significant political differences between sample and omitted But some (expected) demographic/regional differences

9 Using the data When and where do parties/candidates discuss their Negative message = Talking about an opponent (Lau and Pomper, 2002) Parties did a lot of talking about each other in 2015

10 Leaflets containing at least one negative message
Note: Each leaflet is weighted by the total no. of leaflets collected in the constituency.

11 Targets of negative messages – Major parties

12 Targets of negative messages – Minor parties

13 Using a selection model to understand when and where parties talk about their opponents
1st stage: Seat included 2010 Margin of victory (-) 2010 Incumbent party Constituency affluence (+) Population density (+) Internet take-up (+) 2nd stage: Leaflet contains message about opponent 2010 Margin of victory (-) Incumbent (-) Female candidate (-) Black or ethnic minority candidate Author party

14 First stage: Leaflet from seat included
 Variables Coef. Std. err. 2010 Margin of victory -0.008** (0.003) 2010 constituency status (ref=Other-hold) Conservative-hold 0.631** (0.281) Labour-hold 0.524** (0.289) Lib Dem-hold 1.012** (0.299) Constituency affluence 0.328** (0.047) Population density 0.010** (0.002) Internet take-up -0.291** (0.218) Constant 1.793** (0.302) N (Censored N) 3342 (199) ρ -1.426** Wald Χ2 [p > Χ2] 15.35 [0.0001] 

15 Second stage: Leaflet contains message about opponent
 Variables Coef. Std. err. Predicted change 2010 Margin of victory -0.005** (0.002) +0.074** Incumbent -0.130** (0.062) -0.037** Female candidate (ref=male) -0.171** (0.053) -0.050** BME candidate (ref=white) -0.119** (0.085) -0.034** Party (ref=Labour) Conservative (0.071) -0.014** Lib Dem 0.075** (0.072) +0.018** SNP -0.537** (0.137) -0.175** Green -1.227** (0.088) -0.445** UKIP -1.316** (0.081) -0.478**

16 Why do we care? Every party leader promised a positive campaign
But we have few ways of evaluating how well promises were kept Objective measures of negativity are important! UKIP perceived to have the most negative campaign (YouGov) But perceptions suffer from partisan bias (Bartels, 2002) Election communications are critical in British elections But we know very little about what is in them

17 A new dataset for studying electoral communications
No other data available that allow us to explore systematic variation in messages But, some limitations Self-reported data No knowledge of the population of leaflets Missing areas Thoughts on this are most welcome! Findings consistent with American politics, (limited) previous research in Britain

18


Download ppt "Caitlin Milazzo University of Nottingham Jesse Hammond"

Similar presentations


Ads by Google