Brixtofte-skandalens effekt på den offentlige opinion

For snart tyve år siden, onsdag den 6. februar 2002, kunne B.T. rapportere, at Farum-borgmester Peter Brixtofte, der også var medlem af Folketinget for Venstre, havde haft et ekstremt forbrug af luksusgoder, herunder et forbrug af meget dyr rødvin. Hele gildet var som bekendt betalt af skatteborgerne.

Påvirker politiske skandaler af den karakter borgernes opfattelse af politikerne? En lang række af studier har undersøgt, hvad der blandt andet påvirker borgernes opfattelse af korruption (for en god introduktion til denne litteratur, se Blais et al. 2015), men der er relativt få analyser, der belyser, hvordan politiske skandaler påvirker sådanne generelle opfattelser – især i en dansk kontekst (og altså ikke kun om borgernes holdninger til de skandaleramte politikere ændrer sig).

I dette indlæg anvender jeg et kvasi-eksperimentelt forskningsdesign med data fra Valgundersøgelsen 2001, indsamlet i dagene omkring Brixtofte-skandalen, til at undersøge, hvordan danskernes holdninger blev påvirket – eller ikke påvirket – af skandalen. Disse data blev indsamlet i den periode, hvor Peter Brixtofte ramte forsiden på B.T. og skandalen brød ud i lys lue i hvad der må siges at være den største kommunalpolitiske skandale i dansk politik.

Politiske skandaler, med deres fokus på kontroverser og fejltagelser, tager en stor del af mediernes opmærksomhed, når de finder sted (Fogarty 2013, Nyhan 2015, Puglisi og Snyder 2011). Politiske skandaler giver således ofte borgerne utvetydig information omkring, hvad politikere har gjort forkert. Derfor er det også forventeligt, at borgere er mere tilbøjelige til at opfatte politikerne som værende mere korrupte og have mindre tillid til poltikerne ovenpå Brixtofte-skandalen.

Det er vigtigt at forstå hvordan og hvorfor sådanne skandaler påvirker borgernes opfattelse af politikerne, især i forhold til opfattelser af korruption blandt og tilliden til politikerne (Bowler og Karp 2004, Maier 2011). Den akademiske litteratur, der studerer hvordan borgerne reagerer på skandaler, har blandt andet undersøgt skandalers betydning for borgernes opfattelser af politikernes personlighedstræk (Funk 1996), tilliden til politikerne (Bhatti et al. 2013) og den generelle støtte til politikerne (Alford et al. 1994, Stoker 1993). Fælles for mange af disse studier er, at de kigger på de politikere, der er direkte involveret i politiske skandaler og ikke vælgerens opfattelse af politikerne i almindelighed.

Et generelt fund i litteraturen er, at skandaler har negative effekter på den offentlige opinion. Kumlin og Esaiasson (2012) viser i et komparativt studie at borgernes tilfredshed med demokratiet falder, når politiske skandaler finder sted. Bowler og Karp (2004) finder at borgere, der oplever en politisk skandale, har lavere tillid til politikerne og politiske institutioner. Solé-Ollé og Sorribas-Navarro (2018) undersøgte hvordan korruptionsskandaler påvirkede holdningerne til politikerne i Spanien og fandt, at sådanne skandaler påvirkede både borgernes tillid til politikerne såvel som deres opfattelse af hvor korrupte politikerne er. Ares og Hernández (2017) fandt eksempelvis at en korruptionsskandale i Spanien påvirkede borgerens tillid til politikerne.

Brixtofte-skandalen

For en fantastisk gennemgang af skandalen, der er både lærerig og underholdende, og med en udførlig beskrivelse af forløbet, kan jeg varmt anbefale Morten Pihl og Jakob Priess-Sørensens bog Brixtofte – Historien om en afsløring fra 2002. En gennemgang af forløbet gives også i dokumentarserien Brixtofte – helt, skurk, far.

Nedenstående forside fra B.T. fra den 6. februar 2002 viser, at Peter Brixtofte havde drukket for 150.000,- på ét døgn. Der var i dækningen i de efterfølgende uger talrige eksempler på borgmesterens magtmisbrug, og disse beskyldninger om magtmisbrug tog kun til i styrke i løbet af de efterfølgende uger og måneder.

Den konkrete skandale er ideel til at undersøge effekten af politiske skandaler på den offentlige opinion af mindst tre grunde. For det første tillader skandalen at operere med en klar sondring mellem før og efter skandalen rammer forsiderne. For mange andre politiske skandaler sker denne udvikling gradvist, hvor det ikke altid er klart, hvornår en politiker eller parti går fra at være i modvind til at være i en egentlig skandale. I dette tilfælde er det klart, at der ikke var en skandale 5. februar og at der var en skandale 6. februar. Som de to journalister Pihl and Priess-Sørensen nævner i deres bog om skandalen, arbejdede de på historien i flere månder, og det var vigtigt for dem at offentligheden først fik nys om skandalen, når de havde tilstrækkeligt med materiale til at køre historien.

For det andet er der tale om en skandale, hvor der ikke var nogen tvivl omkring, hvorvidt der egentlig var tale om forkert adfærd. I mange politiske skandaler er der ikke enighed om, hvorvidt der er tale om en skandale i første instans. Når medierne kunne rapportere at Trump havde gjort noget, der var på kant med loven, var mange klar til at sige, at det var en skandale, hvorimod andre kunne sige, at der ikke var tale om en skandale. Brixtofte-skandalen var ikke alene en skandale fra dag 1, men også en skandale, hvor der ikke var tvivl omkring det faktum, at der var tale om forkert adfærd i form af magtmisbrug.

For det tredje fik skandalen meget omtale i medierne, som ikke var isoleret til et eller få medier. Dette er afgørende da tidligere forskning har vist, hvordan karakteristika ved en skandale kan påvirke, hvor meget opmærksomhed de får i medierne (Puglisi og Snyder 2011). I dette tilfælde fik skandalen opmærksomhed blandt alle aviser såvel som omfattende dækning på nationalt TV og radio, og der var ingen tegn på, at nogle medier ignorerede skandalen – eller såede tvivl om, hvorvidt der var tale om en skandale.

Der var i perioden op til 6. februar 2002 ingen fokus på en stor skandale med Peter Brixtofte i centrum. En sådan dækning ville være problematisk, da det ville indikere at “kontrolgruppen” ikke var en kontrolgruppe, hvilket i værste fald ville give forkerte estimater i analyserne. Nedenstående figur viser antallet af artikler før og efter 6. februar 2002, der nævner ord, der kan være forbundet med et fokus på en skandale. Disse tal er indsamlet via Infomedia. I det første panel ser vi antallet af artikler, der nævner kommunen (“Farum”), borgmesteren (“Brixtofte”) og borgmesterens parti (“Venstre”). Det andet panel viser antallet af artikler, der nævner borgmesteren og “skandale”. Det tredje panel viser antallet af artikler, der foruden at nævne borgmesteren og skandale også nævner “Venstre”.

Der er et begrænset fokus på borgmesteren i medierne fra 15. december 2001 til 6. februar 2002. Der var kun 37 artikler i denne periode. Antallet af artikler der inkluderede Brixtofte og skandale var minimal, og ingen af disse artikler indikerede en politisk skandale a la det man ser fra 6. februar og frem. Tværtimod ser vi et stort fokus på borgmesteren og skandale efter den 6. februar. Dette gør, at vi dermed kan sammenligne borgernes holdninger lige før og efter skandalen til at belyse, hvilken effekt skandalen havde på den offentlige opinion. Jo mere borgerens holdninger er identiske før og efter skandalen, desto mindre indflydelse vil skandalen have haft.

Metode og data

Skandalen fandt sted midt i dataindsamlingen til Valgundersøgelsen 2001. Folketingsvalget i 2001 fandt sted 20. november, men grundet en forsinkelse blev det meste af dataindsamlingen foretaget i 2002. Dataindsamlingen blev foretaget af Gallup i form af CAPI-interviews. Det første interview blev foretaget 15. december 2001. Kun 35 interviews (1,7%) blev foretaget i 2001, og de resterende blev foretaget i perioden fra 1. januar til 14. marts 2002. Dette giver os en unik mulighed til at konstruere to grupper, der er sammenlignelige med en væsentlig undtagelse: deres eksponering til Brixtofte-skandalen. Den ene gruppe består af de respondenter, der deltog i Valgundersøgelsen før 6. februar 2002, hvor den anden gruppe består af de respondenter, der deltog efter 6. februar. 952 respondenter (47,86%) deltog i undersøgelsen før skandalen, og 1.037 respondenter (52,14%) deltog efter skandalen.

Til at måle respondenternes opfattelse af, hvor korrupte politikerne er, har de besvaret følgende spørgsmål: “Hvor udbredt tror De korruption som f.eks. bestikkelse er iblandt politikere i Danmark?”. Svarmulighederne er “Meget udbredt”, “Ret udbredt”, “Ikke særlig udbredt” og “Det sker næsten aldrig”. Jeg har kodet denne variabel, så højere værdier betyder, at respondenten mener, at korruption er mere udbredt blandt politikerne i Danmark. Til at måle borgernes tillid til politikerne, anvender jeg følgende spørgsmål: “Hvor stor tillid har De til danske politikere i almindelighed? Har De meget stor tillid, ret stor tillid, ret lille tillid, eller meget lille tillid?”. Her er svarmulighederne de fire nævnte muligheder.

Fordelen ved dette forskningsdesign er, at det tager udgangspunkt i en “virkelig” skandale og ikke opdigtede scenarier, der kan være en udfordring for den eksterne validitet (Maier 2011). Som Bhatti et al. (2013: 408f) korrekt har pointeret: “the nature of scandals (one-time events that affect all voters simultaneously) render them difficult to study using conventional observational methods (for example, cross-sectional surveys)” (for en lignende pointe, se Blais et al. 2015). Min tilgang her giver således mulighed for rent faktisk at bruge tværsnitsdata til at generere to tilnærmelsesvist identitiske grupper, hvoraf kun den ene er påvirket af den politiske skandale. Med andre ord giver Valgundersøgelsen 2001 en ganske unik mulighed for at undersøge effekten af politiske skandaler på den offentlige opinion.

Nedenstående figur viser hvornår respondenterne deltog i valgundersøgelsen. Den røde stiplede linje angiver den 6. februar 2002, hvor B.T. bragte den første forsidehistorie med Peter Brixtofte.

Brixtofte-skandalen påvirkede danskernes opfattelser

Til at analysere effekten af Brixtofte-skandalen på den offentlige opinion, i.e. borgerens tillid til politikerne og deres opfattelse af korruption blandt samme, estimerer jeg fire regressionsmodeller. Mere specifikt estimerer jeg to regressioner til hver afhængig variabel, en uden kontrolvariable og en med kontrolvariable. De inkluderede kontrolvariable er køn, alder, uddannelse, indkomst, om man er gift, hvor stor en by man bor i og om man stemte på et rødt eller blåt parti ved folketingsvalget den 20. november 2001.

Nedenstående figur viser resultaterne fra de fire regressionsmodeller. Konkret viser figuren effektestimaterne med 95% konfidensintervaller. Her kan vi se, at borgernes tillid til politikerne ikke ændrede sig ovenpå skandalen, men at de omvendt var mere tilbøjelige til at opfatte politikerne som korrupte.

Resultaterne bekræfter at politiske skandaler kan påvirke borgernes generelle opfattelse af, hvor korrupte politikerne er (Bowler og Karp 2004). Det er således ikke kun de involverede politikere, der bliver påvirket af en skandale. Når det er sagt viser analyserne også, at det ikke er på alle dimensioner, at borgerne ændrer deres holdninger. Konkret var vælgerne ikke mere tilbøjelige til at have mindre tillid til politikerne generelt blot fordi én borgmester var involveret i en skandale.

Nogle af studierne angivet ovenfor viser, at politiske skandaler kan påvirke både borgernes tillid til politikerne og deres opfattelse af korruption blandt samme. Der er dog også eksempler på studier, der ikke finder sådanne effekter. Maier (2011) viste eksempelvis i et eksperiment, at borgernes tillid til politikerne, deres tilfredshed med demokratiet og opfatelse af korruption blandt politikerne ikke var påvirket af eksponeringen til en politisk skandale. Dette kan således indikere, at vi skal være påpasselige med at generalisere fra én skandale til politiske skandaler generelt.

Begrænsninger

Der er mindst tre nævneværdige begrænsninger ved ovenstående analyse.

  1. Nutidige skandaler. Meget har ændret sig på 20 år, og der er ingen garanti for, at en lignende skandale i dag ville have tilsvarende effekter. Den måde hvorpå politiske begivenheder dækkes på i dag afviger signifikant fra, hvordan Brixtofte-skandalen blev dækket. Det er svært at sige, hvordan skandalen ville have været dækket, hvis Facebook og Twitter fandtes i 2002.
  2. Forskellige skandaler. Som angivet ovenfor, skal vi være påpasselige med at konkludere for meget på baggrund af én skandale med bestemte karakteristika. Ikke alle politiske skandaler er ens, og forskellige typer af skandaler (Basinger 2013) såvel som karakteristika ved de involverede politikere (Berinsky et al. 2011, Smith et al. 2005) kan have betydning for, om og hvordan borgerne reagerer på en politisk skandale.
  3. Partipolitik. Der var ingen partipolitiske konflikter omkring skandalen, da den stod på. Venstre gjorde det entydigt klart på det nationale niveau, da skandalen tog fart, at der var tale om et lokalpolitisk anliggende. Hvis partiet havde ageret anderledes, er det også muligt, at vi havde set andre effekter. Tidligere forskning har eksempelvis vist, at hvordan politikerne diskuterer en skandale påvirker effekten af selvsamme (Woessner 2005).

Konklusion

Politiske skandaler kan påvirke den offentlige opinion. Analyserne præsenteret ovenfor indikerer, at Brixtofte-skandalen øgede danskernes opfattelse af korruption blandt politikerne generelt, men ikke havde nogen betydning for borgernes tillid til politikerne.

Det er en sag der primært har historisk interesse, men ikke desto mindre en utroligt interessant case, der stadig refereres til den dag i dag. Den gode nyhed er, at skandalen – i hvert fald på kort sigt – ikke rykkede ved vælgernes generelle tillid til politikerne.

Meningsmålinger og demokratisk indflydelse

Det er relativt begrænset hvor stor indflydelse din stemme har ved et folketingsvalg. Det er af samme grund teorier omkring rationel vælgeradfærd konkluderer, at det – i langt de fleste tilfælde – ikke er rationelt at stemme.

Heldigvis er der mange måder at øge sin politiske indflydelse i Danmark anno 2021. Foruden det at stemme ved forskellige valg kan man skrive læserbreve, deltage i vælgermøder, melde sig ind i et politisk parti, donere penge til partier og andre organisationer, kontakte politikere, deltage i underskriftsindsamlinger, like og kommentere på sociale medier og så videre.

Hvis man som ordinær borger – uden mange penge på lommen – vil øge sin indflydelse, vil jeg argumentere for, at den bedste strategi er at deltage i meningsmålinger. Grunden til dette er, at de politiske partier i dag får foretaget meningsmålinger for at finde ud af, hvilke politiske tiltag der er opbakning til i befolkningen. Især når det vedrører større politiske tiltag og reformer, der kan være kontroversielle.

Det er især regeringspartier, eller partier med ønsket om at få regeringsmagten på et tidspunkt, der laver denne slags meningsmålinger. Og det er især med henblik på at vide, hvad vælgere omkring midten synes. Det vil ikke være nogen overraskelse, at folk der stemmer på Enhedslisten er modstandere af en lavere dagpengesats, eller at folk der stemmer på Dansk Folkeparti er tilhængere af yderligere asylstramninger. Stemmer du omvendt på eksempelvis Socialdemokratiet eller Venstre, kan du have relativt stor indflydelse i en meningsmåling, der kan være med til at forme de politiske partiers politik. Især hvis du ikke er loyal partisoldat.

Ligeledes vil du have større indflydelse i en meningsmåling, hvis du repræsenterer en sociodemografisk gruppe, der normalt ikke deltager i meningsmålinger. Jo mere atypisk det er, at din sociodemografiske gruppe vil deltage i en meningsmålinger, desto større vægt vil der blive givet til dit svar (under forudsætning af, at den er repræsentativ for, hvad andre fra din sociodemografiske gruppe, vil mene).

Min pointe er således den simple, at din “stemme” i en meningsmåling vil betyde relativt meget, især hvis man sammenligner det med andre typer af demokratisk deltagelse. Så hvis du vil øge din politiske indflydelse, bør du deltage i meningsmålinger.

25 interesting facts #12

276. People overestimate their own IQ by around 30 IQ points and their partner’s IQ by almost 40 IQ points (Gignac and Zajenkowski 2019)

277. Democratic satisfaction drops significantly when early elections are called by prime ministers or presidents (Morgan-Jones and Loveless 2021)

278. Smoke alarms for children are more effective if they have an maternal voice alarm instead of a tone alarm (Smith et al. 2018)

279. The emotional arcs of stories are dominated by six basic shapes (Reagan et al. 2016)

280. People have a strong counterfactual curiosity (FitzGibbon et al. 2021)

281. In the United States, while conservatives report greater needs for certainty than liberals, these needs are not a major source of political bias (Guay and Johnston 2021)

282. Taxi drivers take longer routes on metered airport routes than Uber drivers, especially for nonlocal passengers (Liu et al. 2021)

283. Employees who are financially dependent on tips are more likely to experience customer sexual harassment (Kundro et al. 2021)

284. Nudging people to compare pigs to dogs make them less permissible of eating animals (Horne et al. 2021)

285. Honey bees understand the concept of zero (Howard et al. 2018)

286. There was European presence in the Americas in 1021 (Kuitems et al. 2021)

287. In the United States, the introduction and marketing of OxyContin explain a substantial share of overdose deaths (Alpert et al. 2021)

288. The Prohibition in Chicago consolidated the organizational elites and increased gender inequality in organized crime (Smith 2020)

289. Sleep inspires insightful behaviour and creativity (Cai et al. 2009; Wagner et al. 2004)

290. Hosting refugees does not increase the likelihood of new conflict, prolong existing conflict, or raise the number of violent events or casualties (Zhou and Shaver 2021)

291. On Wikipedia, biographies about women are more frequently considered non-notable and nominated for deletion compared to men’s biographies (Tripodi 2021)

292. The very wealthy have been constructing luxury basements across certain parts of London, totaling an estimated excavated depth of 25,461 meters since 2008 (Burrows et al. 2021)

293. Based on the movies, James Bond would be in danger because of tropical diseases and diarrhoea due to poor hygeine (Graumans et al. 2021)

294. There is greater than 99% consensus on human caused climate change in the peer-reviewed scientific literature (Lynas et al. 2021)

295. Dogs can display jealous behaviour (Bastos et al. 2021)

296. There is no evidence for systematic voter fraud in the 2020 US presidential election (Eggers et al. 2021)

297. The ability to name unrelated words is a strong measure of creativity (Olson et al. 2021)

298. Spontaneous face touching is an ubiquitous behavior that occurs in people up to 800 times a day (Spille et al. 2021)

299. There is no consistent effect of testosterone on economic risk-taking (Stanton et al. 2021)

300. Nuclear power’s contribution to climate change mitigation is and will be very limited (Muellner et al. 2021)


Previous posts: #11 #10 #9 #8 #7 #6 #5 #4 #3 #2 #1

The political implications of climate change

What are the political implications of climate change? I have been looking into a few studies lately on the implications of climate change for societies. In brief, there is substantial evidence that climate change matter for economic development, political instability, climate migrants, public health, conflict, etc. Nerini et al. (2019) provide a great overview on how climate change can impact the achievement of the different Sustainable Development Goals (for each of the targets in the 17 SDGs):

They find that climate change can undermine the achievement of at least 16 out of the 17 SDGs. In other words, it is clear that climate change is not unrelated to other challenges faced by societies today (I wrote about this in relation to COVID-19 earlier this year). For topics such as poverty, gender equality, and economic inequality, climate change is important.

A lot of the studies I have found look at how climate change shape economic outcomes. In the table below I show some of these studies with information on their climate focus as well as the key result. The core finding is that changes related to climate change (extreme weather events, temperature increase) matter for economic outcomes.

Study Climate focus Result
Coronese et al. (2019) Extreme natural disasters (floods, extreem temperatures, droughts, storms, wildfires, and landslides) Increase in economic damages (GDP)
Dell et al. (2012) Average temperature Decrease in economic growth
Estrada et al. (2017) Global temperature Increase in economic costs (% GDP)
Hsiang et al. (2017) Global mean temperature Increase in economic market and nonmarket damage
Kotz et al. (2021) Increase in day-to-day temperature variability Reduction in regional growth rates
Pretis et al. (2019) Global mean surface temperature Decrease in economic growth

The studies are particularly interested in economic growth rates and economic damages caused by temperature changes.

Similarly, there are studies demonstrating how climate change matters for (political) conflicts. In the table below I provide an overview of some key studies in the field.

Study Climate focus Result
Burkhardt et al. (2019) Air pollution exposure Increase in violent crimes
De Juan and Wegenast (2020) Average annual temperature Decrease in food riot occurrence
Harp and Karnauskas (2020) Surface air temperature Increase in violent crimes
Hsiang et al. (2013) Warmer temperatures and more extreme rainfall Increase in human conflict (interpersonal violence, political instability etc.)
Jun and Sethi (2021) Extreme weather events Increase in military conflicts
Schleussner et al. (2016) Climate-related disasters Increase in risk for armed conflict

These studies tend to focus on temperature and extreme weather events as well, but with at least one study looking at air pollution exposure. Given the different measures available in the different studies, I would be interested in knowing how many of these studies replicate if they used different measures.

What I find most interesting is not the evidence I have found, but rather what I have not found. Of course, I might have missed some relevant studies beyond economic outcomes and (violent) conflict, but I was expecting to see a lot more political science research on the impact of climate change. How will climate change – and not just the framing of climate change or the amount of rain on Election Day – matter for politics?

As climate change is not affecting countries in a homogeneous manner, there should be a lot of options to study the political implications of climate change. However, again, what are the political implications of climate change? We don’t really know. That being said, I do expect that we will see a lot more research on this topic in the coming years. To be continued…

Hvornår kommer folketingsvalget? #2

I et tidligere indlæg argumenterede jeg for, at der var grund til at forvente, at folketingsvalget kunne finde sted samme dag som kommunalvalget. Mit kvalificerede gæt var således, at folketingsvalget ville finde sted 16. november 2021. Vi kan nu konstarere, at jeg tog fejl. Det er ikke første – og helt sikkert ikke sidste – gang, at jeg tager fejl.

Jeg skriver dette indlæg af de samme grunde som jeg aldrig vil kunne blive politisk kommentator, nemlig at jeg finder det lige så relevant at fremhæve, når jeg tager fejl – som når jeg rammer plet. Faktisk finder jeg det mere interessant at bruge tid på at dissekere, hvorfor og hvornår jeg tager fejl.

Man behøves ikke have læst flere bøger af Philip E. Tetlock for at vide, at selv politiske kommentatorer ofte tager fejl. I 2015 viste jeg, at mange politiske kommentatorer tog fejl i forhold til, hvornår folketingsvalget ville blive udskrevet. Det er derfor ikke den store overraskelse, at jeg heller ikke ramte plet i mit kvalificerede gæt denne gang. Det relevante spørgsmål der står tilbage, er, hvorfor vi ikke ser et folketingsvalg nu.

De fleste meningsmålinger er fine for Socialdemokratiet, men partiet ligger ikke trygt på den gode side af de 30% som de gjorde, da jeg skrev mit indlæg. Meget kan ske i en valgkamp og det er ingen selvfølge, at partiet vil stå stærkere på den anden side af et valg. Der har været artikler på det seneste om, at rød blok ikke længere har et stabilt flertal (se eksempelvis dækningen af meningsmålingerne fra Voxmeter og Megafon).

Dertil skal det tilføjes, at der også har været nogle meningsmålinger på kommunalt niveau, der nok heller ikke er gået Socialdemokratiets opmærksomhed forbi, herunder målinger i København, Fredericia, Silkeborg, Bornholm, Hedensted og Horsens, der alle viser en tilbagegang til Socialdemokratiet (omend i varierende størrelse).

Jeg burde i højere grad have taget højde for, at det ikke kræver de store vælgervandringer før et komfortabelt flertal til rød blok kan reduceres til et meget tæt valg. Min generelle opfattelse er, at meningsmålingerne ikke viser store ændringer over kort tid, men det er ikke usandsynligt at et parti kan rykke sig signifikant over en periode på et halvt år. Med det in mente er det intet chok, at målingerne nu viser noget andet end de gjorde, da jeg skrev mit indlæg.

Det kontrafaktiske spørgsmål er her, om Socialdemokratiet ville have udskrevet valg, hvis meningsmålingerne var bedre. Det kan vi i sagens natur ikke vide, men jeg vil stadig argumentere for, at sandsynligheden for at vi ville have set et folketingsvalg snart, alt andet lige, ville være større hvis meningsmålingerne havde givet Socialdemokratiet og især rød blok et solidt forspring.

En anden mulighed er, at det aldrig har været en seriøs overvejelse i regeringstoppen at udskrive valg i 2021, men at det blot har handlet om at signalere muligheden, især for at sikre, at oppositionspartierne har brugt energi på at være klar til et eventuelt efterårsvalg. Det kan vi forhåbentlig få nogle svar på i fremtiden, når den nuværende politiske situation ligger så langt tilbage i tid, at politiske biografier kan belyse, hvilke strategiske overvejelser, der har været på spil.

En tredje mulighed – som dog potentielt hænger sammen med partiets opbakning i meningsmålingerne – er usikkerheden omkring de politiske sager, der har været i løbet af de seneste måneder. Dette være sig eksempelvis sygeplejerskernes strejke og minksagen, der har stået højere på Socialdemokratiets dagsorden end tankerne om at udskrive valg. Ligehedes har flere ministre ikke nødvendigvis vist sig fra deres bedste side i år, herunder blandt andet Astrid Krag, Jeppe Kofod og Trine Bramsen, hvorfor det kan give god mening ikke at udskrive valg i kølvandet på dette.

Med andre ord burde jeg også have taget højde for, at den politiske dagsorden ikke nødvendigvis ville være domineret af corona (som den trods alt var i begyndelsen af i år), men tværtimod mere nuanceret.

Dette besvarer selvfølgelig ikke, om Socialdemokratiet burde udskrive valg nu. Der er ingen garanti for, at opbakningen til regeringen og dets støttepartier vil være større i efteråret ’22 eller foråret ’23. Det må eftertiden vise.

Hvornår vil vi så se et folketingsvalg? Det har jeg ikke længere et kvalificeret bud på. Mine overvejelser herom er trivielle (jo bedre målinger til rød blok, desto større sandsynlighed for et valg), og jeg vil herfra lade de politiske kommentatorer om at sætte en præcis dato på, hvornår vi ser et valg.

Correlation and causation with Friends

There has been a lot of talk about the TV show Friends in 2021, especially related to the reunion of the cast. I have written about the show before (in Danish), and while it is not a show that I am going to watch again anytime soon, there are a few scenes that might be useful for teaching purposes.

Specifically, I used a few clips in the past when I was teaching students about how correlation does not imply causation. The first clip is about a conversation between Joey and Rachel on why their fridge broke down. The second clip is a conversation between most of the friends about what happens when Phoebe visits the dentist.

Of course, the examples are superficial and I have primarily used the two examples in my introdoctory teaching to let students discuss correlation and causality in the simplest of terms. My experience is that both examples work well.

Last, when teaching statistics, you might also consider using data related to Friends. Emil Hvitfeldt has released an R package with the entire transcript from the TV show Friends. The package is on CRAN. There is some good material on how to analyse the data in the TidyTuesday videos from David Robinson and TidyX. Noteworthy, this is not the only R package related to Friends. You also have the centralperk package that enables you to get random quotes from Friends.

If you would like to analyse data from other TV shows, you can also check out the entire transcript from The Office. For examples on how to analyse this data, I can highly recommend the blog posts by Eric Ekholm (part 1, part 2, part 3).

Updating the replication material for “Welfare Retrenchments and Government Support”

In 2017, I pushed the replication material for my article, ‘Welfare Retrenchments and Government Support’, to a GitHub repository. I had been working on the article for years and the code was not necessarily up to date. It worked perfectly, gave the exact estimates and was relatively easy to read. Accordingly, everything was good, life was simple and I felt confident that I would never have to look at the code again.

This turned out not to be the case. I recently got a mail from a student who was unable to get the exact estimates as reported in Table 1 in the paper, even when following my script and using the data I made publicly available. I went through the code and I noticed that I could not reproduce the exact estimates with my current R setup. Sure, the results were substantially identical but not the exact same – and the N was also different.

I looked into the issue and I could see that changes were made to the defaults of set.seed() in R 3.6.0. As I ran the original analyses in R 3.3.1, and I am now using R 4.1.0, this could explain why the matching procedure I rely on is not returning the exact matches. For that reason, I decided to make some updates to the replication material so there now is a dataset with the matched data. The script is doing the same as before, but it is not relying on the matched data obtained with the setup in R 3.3.1. This should make it a lot easier to get the exact same estimates as provided throughout the paper.

To increase the changes of long-term reproducibility, I should consider using packrat or a Docker container (I primarily use Docker for my Shiny dashboards). However, as the analyses are mostly a few OLS regressions, I believe this would be overkill and would not necessarily make it easier for most people to easily download the data and script and play around with the results. And I don’t mind making extra updates in the future if needed in order to reproduce the results with different setups.

Interestingly, I did all of these analyses before I doubled down on tidyverse and for that reason I decided to make a series of additional updates to the material, including:

  • More spaces to make the code easier to read. For example, instead of x=week, y=su it is now x = week, y = su.
  • The use of underscores (snake cases) instead of dots. For example, the object ess.matched is now ess_matched.
  • A significant reduction in the use of dollar signs (primarily by the use of mutate()).
  • The use of pivot_longer() instead of gather().
  • No double mention of the variable edulevel in the variable selection.
  • Removing the deprecated type.dots argument from rdplot().
  • The use of seq(0.01, 0.25, 0.01) instead of having 0.01, 0.02, 0.03, 0.04, etc. all the way to 0.25!
  • The use of map_df() instead of a for loop.

And a series of other minor changes that makes the code easier to read and use in 2021. I have made the updated material available in the GitHub repository. There is a revised R-script for the analysis, a dataset with the matched observations and a file with the session info on the current setup I used to reproduce the results.

I have started using the new native pipe operator in R (|>) instead of the tidyverse pipe (%>%), but I decided not to change this in the current version to make sure that the script is also working well using the version of R I used to conduct the analysis years ago. In other words, the 2021 script should work using both R 3.3.1 and R 4.1.0.

I also thought about using the essurvey package to get the data from the European Social Survey (we have an example on how to do that in the Quantitative Politics with R book), but I find it safer to only work with local copies of the data and not rely on this package being available in the future.

In a parallel universe a more productive version of myself would spend time and energy on more fruitful endeavors than updating the material for an article published years ago. However, I can highly recommend going through old material and see whether and if it still works. Some of the issues you might encounter will help you a lot in ensuring that the replication material you create for future projects are also more likely to stand the test of time.

Social science research during COVID-19 #2

I have been reading a few papers on conspiracy theories in relation to COVID-19 (primarily because I was asked to read a study on this topic for a journal). In parallel, I have been following the debates taking place between scientists on the origin of COVID-19. In this post, I will argue that there are some interesting discprecancies in how different researchers look at and talk about the origin of COVID-19.

More specifically, consider the lab leak theory on the origin of COVID-19. There are several social science studies that treat this as a conspiracy theory. A few examples are in place. More than a year ago, a study published in June 2020 mapped “retweets that attribute COVID-19 to a bioweapon or a lab in China” to examine the “geography behind a metonymic conspiracy theory”. Another study from September 2020, examined “a conspiracy theory that it [COVID-19] was human-engineered and leaked, deliberately or accidentally, from a research laboratory in Wuhan, China.” A third study from December 2020 looked at “popular conspiracy theories around a Chinese laboratory in Wuhan”. And a fourth study from June 2021 considered “the Wuhan laboratory and 5G theories as classic conspiracy theories”. Of course, this is not a comprehensive review but just to illustrate that there are multiple studies on conspiracy theories within the social sciences that treat the lab leak theory as a conspiracy (similar to that of the 5G conspiracy).

What is interesting about the study of the lab leak theory as a conspiracy theory? The fact that people outside the social sciences do not talk about the lab leak theory as a conspiracy theory. In May, a group of researchers wrote in Science that “We must take hypotheses about both natural and laboratory spillovers seriously until we have sufficient data.” For a good overview of the timeline and why a lot of people were quick to call the lab leak theory a conspiracy, see this feature published in BMJ. There is also a good explainer in Nature on what we know, including arguments for and against the lab leak theory. The head of the WHO has said that it was premature to rule out the lab leak theory. This is definitely not to say that the lab leak theory is the most promising theory, especially as there is more support for the theory that COVID-19 is the result of a natural spillover from humans to animals (see also this recent article).

I am not an epidemiologist. I know next to nothing about the outbreak of diseases. Well, for the sake of the argument I am about to make, let us say that I know nothing about the outbreak of diseases. The world is a complicated place, especially when we are faced with a new virus like COVID-19. I cannot say anything about what is true or false with any level of authority. I believe we know a lot about the virus (and by “we” I primarily mean the scientific community which I am by no means a part of), and there is still a lot we do not know.

However, in general, I believe it is limited what social science research conducted during the COVID-19 pandemic can say beyond what we can already infer from research into previous crises (including previous research on conspiracy theories). In other words, we do not need to begin from scratch if we are to understand human behaviour in relation to COVID-19. In fact, I believe it is deadly dangerous when social scientists prioritise speed over science when they want to act on an opportunity promptly. In my previous post on the topic, I wrote about how I found it problematic and unethical when a group of economists tried to convince Americans to worry less about COVID-19 in the beginning of 2020!

To give the researchers the benefits of the doubt, I am willing to assume that they are incompetent rather than malicious. We did not know a lot about COVID-19 in the beginning, and it is only fair to expect that we should not even consider the possibility of a lab leak as anything but a conspiracy theory promoted by Donald J. Trump and the like.

That being said, the problem is that such research is damaging. If people who are prone to believe in conspiracy theories can see that social scientists are treating theories like the lab leak theory as a conspiracy, why not expect the same for the real conspiracy theories like 5G? This is why it is paramount that researchers should be careful with how they categorise conspiracy theories, especially if they want to publish on such theories in the midst of a global pandemic.

New book: Reporting Public Opinion

I am happy to announce the publication of a new book, ‘Reporting Public Opinion: How the Media Turns Boring Polls into Biased News‘, co-authored with Zoltán Fazekas. The book is about how and why opinion polls are more likely to be about change in the news reporting. Specifically, journalists are more likely to pick opinion polls that show changes, even when such changes are within the margin of error, highlight such changes in the reporting – and the public, pundits and politicians are more likely to respond to and share such polls.

Here is the puzzle we address throughout the various chapters: how can most opinion polls show a lot of stability over short periods of time whereas the reporting of opinion polls are dominated by change?

Even for the most hardcore followers of politics, opinion polls are quite boring in and by themselves. In most cases they show nothing new. When we take the margin of error into account, a new opinion poll will most likely show that there is no statistically significant shift in the polls for any of the political parties of interest. And when there is a large change, it is most likely a statistical fluke we should be cautious about. I have over the years written countless posts about such opinion polls being covered in the Danish media.

The book is our attempt to provide a unified framework to better understand these dynamics in a systematic manner. In the first chapter of the book, we introduce the theoretical puzzle and outline the main limitation of existing studies on the topic, namely that studies on opinion polls tend to focus on one specific stage in the coverage, such as whether methodological details are present in the coverage or not. To fully understand how opinion polls are covered and consumed in contemporary democracies, we argue that we need to combine different literatures on opinion polls and examine how a strong preference for change can explain biases in how opinion polls travel through several stages from their initial collection to how they reach the public.

In the second chapter, we further develop a framework that focuses on the temporal dimension of how opinion polls are brought to the public via the media. This chapter serves as an introduction to the four stages that opinion polls have to go through in our framework. Specifically, we show how each stage – or activity – will lead to polls showing greater changes getting more attention. This is illustrated below:

Next, throughout Chapters 3, 4, and 5, we cover the stages of opinion polls in greater detail and show collectively how opinion polls are being turned into specific news stories. In Chapter 3, we focus on the selection of opinion polls. That is, we investigate what can explain whether journalists decide to cover an opinion poll or not. In Chapter 4, we target the content of the reporting of opinion polls, which covers the news articles dedicated to the opinion polls that journalists have decided to report on. In doing this, we show how the selection and reporting of opinion polls are shaped by a similar preference for change. Noteworthy, when introducing the idea of change, we dedicate extensive considerations to how we can best measure change and what the availability of these change measures means for the selection and reporting.

In Chapter 5, we analyse the next natural stage in the life of opinion polls: how do politicians, experts and the public respond to them and to the stories written about them. Essentially, we delve into the implications of how these opinion polls are selected and covered. Here, we show that both elites and the broader public have a strong preference to engage with (respond to or share) opinion polls that show greater changes or support a well-defined change narrative. Interestingly, we find that opinion polls showing greater changes are much more likely to go viral on Twitter.

In Chapter 6, we turn our attention to the alternatives of the reporting of opinion polls. Here, we discuss how no opinion polls at all, poll aggregators, social media, and vox pops can be seen as alternatives to opinion polls, and in particular what are their strengths and limitations. The ambition here is not to force the reader to decide whether opinion polls are good or bad, but rather to understand how alternatives to opinion polls can mitigate or amplify the biases introduced in the previous chapter.

Last, in Chapter 7, we conclude how the media might report on opinion polls by considering the trade-offs between what the polls often show and what journalists wish they showed. Specifically, we first set out to discuss the implications of the findings for how we understand the political coverage of opinion polls today and then discuss the most important questions to be answered in future work.

The book is the product of years of work on the topic of how opinion polls are reported in the media. However, while the topic should be of interest to most people with an interest in politics and opinion polls, this is an academic book and I should emphasise that it might be a tough read for a non-academic audience.

You can buy the book at Waterstones, Bookshop, Springer, Blackwell’s and Palgrave.

Causality models: Campbell, Rubin and Pearl

In political science, the predominant way to discuss causality is in relation to experiments and counterfactuals (within the potential outcomes framework). However, we also use concepts such as internal and external validity and sometimes we use arrows to show how different concepts are connected. When I was introduced to causality, it was on a PowerPoint slide with the symbol X, a rightwards arrow, and the symbol Y, together with a few bullet points on the specific criteria that should be met before we can say that a relationship is causal (inspired by John Gerring’s criterial approach; see, e.g., Gerring 2005).

Importantly, there are multiple models we can consider when we want to discuss causality. In brief, there are three popular causality models today: 1) the Campbell model (focusing on threats to validity), 2) the Rubin model (focusing on potential outcomes), and 3) the Pearl model (focusing on directed acyclic graphs). The names of the models are based on the names of the researchers who have been instrumental in the development of these models (Donald Campbell, Donald Rubin and Judea Pearl). I believe a good understanding of these three models is a prerequisite to be able to discuss causal inference within quantitative social science.

Luckily, we have good introductions to the three frameworks that compare the main similarities and differences. The special issue introduced by Maxwell (2010) focuses on two of the frameworks, namely the frameworks related to Campbell and Rubin. What is great about the special issue is that it focuses on important differences between the two frameworks but also how the two frameworks are complementary. That being said, it does not pay a lot of attention to the Pearl’s framework. Shadish (2010) and West and Thoemmes (2010) provide comparisons of the work by Campbell and Rubin on causal inference. Rubin (2010) and Imbens (2010) further provide some additional reflections on the causal models from their own perspectives.

The best primer to understand the three frameworks is the book chapter by Shadish and Sullivan (2012). They make it clear that all three models to causality acknowledge the importance of manipulable causes and brings an experimental terminology into observational research. In addition, they highlight the importance of assumptions (as causal inference without assumptions is impossible). Unfortunately, they do not summarise the key similarities and differences between the models in a table. For that reason, I decided to create the table below to provide a brief overview of the three models. Keep in mind that the table provides a simplified comparison and there are important nuances that you will only fully understand by consulting the relevant literature.

Campbell Rubin Pearl
Core Validity typology and the associated threats to validity Precise conceptualization of causal inference Directed acyclic graphs (DAGs)
Goal Create a generalized causal theory Define an effect clearly and precisely State the conditions under which a given DAG can support a causal inference
Fields of development Psychology Statistics, program evaluation Artificial intelligence, machine learning
Examples of main concepts Internal validity, external validity, statistical conclusion validity, construct validity Potential outcomes, causal effect, stable-unit-treatment-value assumption Node, edge, collider, d-seperation, back-door criterion, do(x) operator
Definition of effect Difference between counterfactuals Difference between potential outcomes The space of probability distributions on Y using the do(x) operator
Causal generalisation Meta-analysis, construct and external validity Response surface analysis, meditational modeling Specified within the DAG
Assumption for valid inference in observational research Ruled out all threats to validity Strong ignorability Correct DAG
Examples of application Quasi-experiments Missing data imputation, propensity scores Mediational paths
Conceptual and philosophical scope Wide-ranging Narrow, formal statistical model Narrow, formal statistical model
Emphasis Descriptive causation Descriptive causation Explanatory causation
Preference for randomized experiments Yes Yes No
Focus on effect or mechanism Effect Effect Mechanism
Limitation General lack of quantification, no formal statistical model (lacks analytic sophistication) Limited focus on features of research designs with observational data Vulnerability to misspecification

The Campbell model focuses on validity, i.e., the quality of the conclusions you can make based on your research. The four types of validity to consider here are: 1) (statistical) conclusion validity, internal validity, construct validity, and external validity. Most important for the causal model is the internal validity. That is, the extent to which the research design identities a causal relationship. External validity refers to teh extent to which we can generalise the causal relationship to other populations/contexts. I believe one of the key advantages here is the comprehensive list of potential threats to validity listed in this work. Some of these potential threats are more relevant for specific designs or results, and being familiar with these potential threats will make you a much more critical (and thereby better) researcher. The best comprehensive introduction to the Campbell model is Shadish et al. (2002).

The Rubin model focuses on potential outcomes and how units have potential outcomes in different conditions (most often with and without a binary treatment). For example, Y(1) is an array of potential outcomes under treatment 1 and Y(0) is an array of potential outcomes under treatment 0. This is especially useful when considering an experiment and how randomisation can realise one potential outcome for a unit that can, in combination with other units, be used to calculate the average treatment effect (as we cannot estimate individual-level causal effects). To solve the fundamental problem of causal inference (that we can only observe one unit in one world) we would need a time machine, and in the absence of such science fiction tools, we are left with the importance of the assignment mechanism for causal inference (to estimate effects such as ATE, LATE, PATE, ATT, ATC, and ITT). One of the key advantages of this model is to understand how potential outcomes are turned into one realised outcome and the assumptions we rely on. For example, the Stable Unit Treatment Value Assumption (SUTVA) implies that potential outcomes for one unit are unaffected by the treatment of another unit. This emphasises the importance of minimising the interference between units. The best comprehensive introduction to the Rubin model is Imbens and Rubin (2015).

The Pearl model provides causal identification through directed acylic graphs (DAGs), i.e., how conditioning on a variable along a path blocks the path, and how specific effects need to be restricted in order to make causal inferences. When using with this model of causality, you are often worken with multiple paths and not a simple setup where you only have two groups, one outcome and a single treatment. DAGs can also be understood as non-parametric structural equation models, and are particular useful when working with conditional probabilities and Bayes networks/graphical models.

One of the main advantages of the Pearl model is that it forces you to think much more carefully about your causal model, including what not to control for. For that reason, the model is much better geared to causal inference in complicated settings than, say, the Rubin model.

However, there are also some noteworthy limitations. Interactions and effect heterogeneity are implied in the model, and it can be difficult to convey such ideas (whereas it is easier to consider conditional average treatment effects in the Rubin model). While DAGs are helpful to understand complex causal models, it is often less helpful when we have to consider the parametric assumptions we need to estimate causal effects in practice.

The best introduction to the Pearl model is, surprisingly, not the work by Pearl himself (although I did enjoy The Book of Why). As a political scientist (or a social scientist more generally), I find introductions such as Morgan and Winship (2014), Elwert (2013), Elwert and Winship (2014), Dablander (2020), and Rohrer (2018) much more accessible.

(For Danish readers, you can also check out my lecture slides from 2016 on the Rubin model, the Campbell model and the Pearl model. I also made a different version of the table presented above in Danish that you can find here.)

In political science, researchers have mostly relied on the work by Rubin and Campbell, and less so on the work by Pearl. However, recently we have seen some good work that relies on the insights provided by DAGs. Great examples include the work on racially biased policing in the U.S. (see Knox et al. 2020) and the the work on estimating controlled direct effects (Acharya et al. 2016).

Imbens (2020) provides a good and critical discussion of DAGs in relation to the Rubin model (in favour of the potential outcomes over DAGs as the preferred model to causality within the social sciences). Matthay and Glymour (2020) show how the threats to internal, external, construct and statistical conclusion validity can be presented as DAGs. Lundberg et al. (2021) show how both potential outcomes and DAGs can be used to outline the identification assumptions linking a theoretical estimand to an empirical estimand. This is amazing work and everybody with an interest in strong causal inference connecting statistical evidence to theory should read it.

My opiniated take is that the three models work well together but not necessarily at the same time when thinking about theories, research designs and data. Specifically, I prefer Pearl → Rubin → Campbell. First, use Pearl to outline the causal model (with a particular focus on what not to include). Second use Rubin to focus on the causal estimand of interest, consider different estimators and assumptions (SITA/SUTVA). Third, use Campbell to discuss threats to vality, measurement error, etc.

In sum, the three models are all good to be familiar with if you do quantitative (and even qualitative) social science.