La place du chercheur dans l’evaluation despolitiques publiques

Esther Duflo

MIT and J-PAL

Le Printemps de l’Evaluation–Assemblee Nationale

L’enjeu

Exploring how people feel about economics ecnmy.org/research

16 17

To what extent do we trust economists?

1 Jan 2017 (UK Supplement) = 24% trust (32% in 2017, 36% 2016)2 ‘Trust in each political leader to communicate honestly - Hammond - 18% of general population and 23% of informed public. The Edelman Trust Index’s ‘Informed public’ group is composed of those who are ‘high income, college educated or follow business and public policy information several times a week or more’3 https://yougov.co.uk/news/2017/02/17/leave-voters-are-less-likely-trust-any-experts-eve/

Answer: we don’t.

‘Globally, almost 50 percent of university-ed-ucated, news-following members of society say the system is not working. In the UK, just 11 percent of the population think the system is working. 29% (3 in 10) are uncertain. 60% believe that the system is failing them. We recorded a strong sense of injustice, a lack of hope and a desire for change.’ Edelman Trust Barometer 2017

The Edelman Trust Barometer (ETB) tracks trust in institutions globally. In January 2017, the barometer found that trust in the British gov-ernment, already as low as 36% at the start of 2016, had fallen to just 26%. The UK’s trust in the media also fell from 36% in 2016 to 24%1.

Post-Brexit, the ETB found that 8 in 10 Brits do not trust political leaders ‘to do what is right’ and that only 18% of the population trust in the Chancellor of the Exchequer (Philip Hammond)

to communicate honestly2.

In a 2017 YouGov poll, 79% stated they did not trust politicians when they talk about their field of expertise – only 5% did3.

A distrust of economics and economists seems to stand out among distrust in institutions as a whole. Our own YouGov polling has shown that only 4% of the UK population felt that discussion and information they heard in the news about the economy was ‘completely reliable and trustwor-thy’, dropping to 1% of respondents in Scotland. The average level of trust in economic discussion and reliability of information about the economy in the population was just 4.4/10.

In Scotland, trust is particularly low; nearly one fifth of respondents felt that discussion and information about the economy was ‘not at all’ reliable and trustworthy. 67% rated trustworthi-ness 5 or under, while only 15% of respondents rated trustworthiness over 7 out of 10.

Economists were the group of ‘experts’ with the single biggest drop in trust for Leave voters. In a recent YouGov poll, only 14% of Leave voters stated they trust economists. 56% stated they

don’t. Of Remain voters, 35% state they trust economists and 36% that they don’t4.

Voting patterns in the EU referendum un-derline polarisations in attitudes towards the effect of specific issues on the economy on the basis of age and education. ‘The young and highly educated are more likely to believe that immigration was good for the economy [and more likely to vote Remain], while older peo-ple and non-graduates were more likely to say it was bad [and more likely to vote Leave].’ British Social Attitudes Survey.

Finally, many have picked up a growing feeling of ‘anti-politics’ (for example: Clarke et al., 2016) that encompasses all or part rejection of current politics and parties, scandal-hit institutions, ‘political correctness’, and/or an ‘elite’ or ‘estab-lishment’.

‘To rebuild trust and restore faith in the system, institutions must step outside of their traditional roles and work toward a new, more integrated operating model that puts people — and the addressing of their fears — at the center of everything they do.’ Foreword, Edelman Trust Index 2017

4 YouGov survey on how far the British public trust various different types of experts when they talk about their own fields of expertise. 2040 people polled, UK representative, 14th-15th Feb 2017. ‘Overall, 44% of UK representative sample state they do not trust economists when talking about their field of expertise. 31% don’t know and 25% trust them.’ - https://yougov.co.uk/news/2017/02/17/leave-voters-are-less-likely-trust-any-experts-eve/

People in Scotland are 5 times more likely to state that they do ‘not at all’ understand what is being said at all than in London, and twice as likely as any other regions in the UK to report this.

Scots are also nearly 3 times more likely to state that they do ‘not at all’ see how economic news affects their day-to-day life and are over twice as likely to state that they do ‘not at all’ feel able to engage in economic conversations to the level they would like to in comparison to the rest of the UK.

La perception des economistes

I Predictions

I Decouvrir les grandes lois de l’economie

I Ingenierie economique.

La valeur des predictions

”The only function of economic forecasting is to makeastrology look respectable” John Kenneth Galbraith

L’evaluation: usages classique

I Ex-post: le programme a-t-il ete mis en place comme prevu :la majeure partie des evaluations en France aujourd’hui.

I Ex-ante: analyse cout-benefice

I Ex-post: quel est l’impact?: a a marche ou a n’a pas marche

La place du chercheur dans ce paradigme

I La montee en puissance de l’evaluation d’impact a donne auxchercheurs une place dans l’evaluation.

I Les chercheurs sont ceux qui maıtrisent ces nouvellesmethodes, qui ont le temps, et qui ont les bonnes incitations.

I Les resultats sont (parfois) utilises apres coup pour avancer,avec des hauts et des bas.

I Evaluation OPP-CVE

OPP-CVE (Behagel, Crepon, Gurgand)

I Une evaluation avec assignation aleatoire de la politiqued’accompagnement renforce des chomeurs de longue duree.

I Une comparaison entre l’implementation par l’operateurpublique (CVE) ou prive (OPP)

I La comparaison montre un meilleur rappport cout beneficepour CVE

I Malgre cela le programme de delegation a d’abord ete etendu.

I Mais un rapport de la Cour des comptes l’a epingle par lasuite, et on est revenu sur la delegation.

Repenser l’une et l’autre

I L’economiste comme plombier

I L’evaluation comme un processus de co-decouverte

L’economiste comme plombier

“If economists could manage to get themselves thoughtof as humble, competent people on a level with dentists,that would be splendid” John Maynard Keynes

L’economiste comme plombier

I Les economistes ont de plus en plus l’occasion d’intervenirdans les details des politiques publiques (encheres publiques,assignation des eleves a l’ecole, problemes de societe)

I Cela demande de se pencher sur les details, qui sontimportants et souvent negliges

I Ce qui demande d’adopter une attitude de plombier: nous neconnaissons pas tous les parametres du probleme a l’avance: ilfaut faire au mieux, ajuster si necessaire.

Assignation des eleves a l’ecole

I Un probleme d’actualite pour la France (parcourssup)

I Qu’on retrouve dans de nombreux pays, de New York et Boston auGhana.

I Al Roth (Prix Nobel) , Parag Pathak (Clark Medal) ont travailleavec des gouvernments, des villes etc. sur ces sujets.

I Il existe de nombreux resultats theoriques sur les algorithmesefficaces pour permettre une assignation des eleves en fonction deleurs preferences sans qu’il y ait besoin d’avoir un comportementstrategique (scoop: il faut demander aux etudiants de ranger leurspreferences dans l’ordre pour eviter les problemes d’affectation).

I Mais il y a de nombreux problemes d’ordre beaucoup plus pratique:I Comment persuader les eleves qu’etre strategique n’est pas

necessaire?I Combien de choix les eleves peuvent raisonablement evaluer?

I Comment creer un systeme qui prenne ces problemes au serieuxsans ignorer le bon sens?

L’evaluation comme co-decouverte

I Le principal interet des evaluations n’est en fait peut-etre pasune meilleure evaluation de l’impact

I C’est le processus lui meme: discipline, reflexion sur leprogramme et ses buts.

I Un programme qui se revele inefficace n’est pas un echec, sion comprend pourquoi il n’a pas marche comme prevu

CV anonymes (Behaghel, Crepon, Le Barbanchon)

I Les employeurs volontaires sont alloues alatoirement a voir(ou pas) le nom du candidat

I L’anonymisation conduit a une baisse du nombre de rappelpour les candidat des minorites

I Un resultat inattendu qui montre que les employeurs(volontaires) essaient en fait de corriger les signaux quipourraient etre negatifs

I Dans la meme veine, la politique “Ban the Box” (quidemande aux candidats s’ils ont un passe criminel) a un effetcontreproductif et conduit a une discrimination plus fortecontre les Noirs.

I La Lutte contre la discrimination est plus compliquee qu’on lepensait.

Conclusion

I L’evaluation d’impact doit rentrer plus largement dans nosmoeurs

I Il ne s’agit pas de tout evaluer (une recette pour ne rienevaluer en fait)

I Mais de saisir chaque opportunite (des programmes innovants,des grands enjeux de societe) pour en apprendre un peu plus achaque fois

I Il est dangereux de penser que nous savons quoi faire et queles reponses sont evidentes

I L’etendue de notre ignorance est immense...

I L’evaluation d’impact ne vient pas se substituer au debat maiselle apporte des elements de debat.