US election modelling & forecasting
features di bastianin
- dependent variable of your regressions, that is the Democratic (D) share of the two party vote defined as 100*D/(D + R)
- real percapita GDP by state
- price index based on GDP deflator defined as nominal (current dollar) GDP divided by real GDP
- a measure of strong growth following the approach of Fair
- for incumbency variables you can use the same variables used in the original paper
further variables such as:
- state demographic characteristics
- economic variables (such as State level unemployment)
- vote-related variables (e.g. the party of the Governor)
- Prezzo Benzina | ultimo anno prezzi petrolio/gas/benzina
featuires di lichtman
- Seggi ai Deputati | Mandato del partito:dopo le elezioni di medio termine, il partito al governo ha più seggi alla Camera dei Rappresentanti di quanti ne avesse dopo le precedenti elezioni di medio termine
- Voti Primarie | Competizione:non esiste una seria competizione per il candidato ufficiale alle primarie di partito
- dummy | Ufficialità: il candidato ufficiale è l'incumben general.
- Percentuale Intenzioni di voto indipendente | Terza parte: Mentre ci sono partiti minori che sono in corsa, come in tutte le elezioni, non ce ne sono che superano il 3% delle intenzioni di voto in qualsiasi sondaggio
- GDP anno | Economia a breve termine: l'economianon è inrecessione durante la campagna elettorale
- GDP mandato | Economia a lungo termine:lacrescita economica pro capite durante il mandato è uguale o superiore alla media degli ultimi due mandati presidenziali
- delta-GDP |
- short-term measures of economic performance:
- Gt growth rate (at an annual rate) of real per capita GDP in the first three quarters of the election year
- long-term measures of economic performance:
- Pt is the absolute value of the inflation rate (at an annual rate) in the first 15 quarters of the administration
- Zt no. of quarters in the first 15 of admin. in which GDP growthexceeded 3.2% at an annual rate
- familiarity of candidate and preference for variety of party in power
- It is 1 if the Democrats (D) are in the White House at the time of the election and –1 if the Republicans (R) are
- DPERt is 1 if D president is running again (D incumbent), –1 if R incumbent and 0 otherwise
- DURt 0 if the incumbent party has been in power for only 1 or 2 consecutive terms, 1 [–1] if D [R] party has been in power for three consecutive terms, 1.25 [–1.25] if the D [R] party in power for 4 consecutive terms, 1.5 [–1.5] if the D [R] party has been in power for 5 consecutive terms, and so on.
- WARt 1 if election year = 1920, 1944, 1948 (0 otherwise)
Vt = β0 + β1(Gt×It) + β2(Pt×It) + β3(Zt×It) + β4(It) + β5(DPERt) + β6(DURt) + ut
Ideas for empirical projects
- Apply ML to Fair’s data to obtain better predictions and correct inferences (Belloni et al., 2014)
- Extend Fair’s work considering electoral votes by US State (also need State level G and P). Also in this case a ML approach that guarantees good predictions and correct inferences is needed (Athey, 2019; Athey and Imbens, 2019)