Congressional Forecasts for 2018: Structure-X Models


We build here on our “Structure-X model,” successfully applied in 2014. We first generate a 2018 forecast from our classic structural model. Next, we adjust this forecast on the basis of expert judgments provided in Inside Elections with Nathan L. Gonzales.

With respect to the first step, the structural model, we employ the following equation:

House Seat Change = Presidential Approvalt–1 + Disposable Incomet–1 + Mid-termt  (Eq. 1)

OLS yields the following results,

HS = -44.83 + 0.82*P + 4.91*I – 28.53*M  (Eq. 2)

R2 = .59, adj. R2 = .55, RMSE = 18.06, D-W = 1.87, N = 35

Notes: HS = presidential party seat change in the House of Representatives, I = change in real disposable income, for initial six months of the election year (from the Bureau of Economic Analysis’s National Income and Product Account Table 2.1: Personal Income and Its Disposition), P = June Gallup poll presidential popularity rating from Gallup’s Presidential Approval Center, M = midterm dummy (0 = presidential election, 1 = midterm election), figures in parentheses are t-scores, * = statistical significance beyond .05.

To forecast the 2018 seat change in the House, we insert the independent variable values: I = 1.73 (December 2017 through June 2018), P = 42 (June 2018), and M = 1:

HS2018 = -44.83 + 0.82 (42) + 4.91 (1.73) – 28.53 (1)  (Eq. 3)

= -30 Republican seats.

As a second step, we now modify that forecast by taking into account expert judgments, using the total “seats in play” number drawn from Inside Elections. For June 2018, Gonzales reports 68 Republican seats in play, versus 10 Democratic seats in play. Thus, the X-number = (10 – 68) = – 58.   This expert index (-58) clearly departs from our structural forecast (-30).  To reconcile these differences, we average the two methods, meaning a June-based forecast of (-30 – 58)/2 = -44 net Republican seat loss.

We also use the same Structure-X modeling strategy to forecast the 2018 Senate midterm election. First, we use our classic Senate structural model to estimate a midterm forecast:

Senate Seat Change = Presidential Approvalt–1 + Disposable Incomet–1 + Mid-termt + Seats Exposedt  (Eq. 4)

OLS yields the following:

SC = 2.54 + 0.14*P + 0.90*I – 2.43*M – 0.69*Expose  (Eq. 5)

R2 = .68, adj. R2 = .64, RMSE = 2.89, D-W = 1.85, N = 35

where SC = presidential party seat change in the U.S. Senate, Expose is number of seats the president’s party has up for reelection, and the other variables are as defined in Equation 2. Inserting the 2018 independent variable values produces the Senate forecast of:

SC2018 = 2.54* + 0.14*(42) + 0.90*(1.73) – 2.43*(1) – 0.69*(9)  (Eq. 6)

= 0 (no seat change)

We adjust this structural model forecast by adding Gonzales’ expert state-level evaluation of races. For July 2018, Gonzales estimates that there are four Democratic party seats in trouble and two Republican seats in trouble. A simple averaging of the structural forecast (zero) and expert judgement (two) produces a Structure-X forecast of a one-seat pickup for the Republicans in the Senate.

Our Structure-X model forecasts that the midterm elections will result in a divided government. The Democratic Party is forecast to win 44 seats. This outcome would give the Democrats a total of 239 House seats and control of the speakership and all House committees. The newly seated House of Representatives controlled by Democrats would be balanced by a Republican Senate and executive branch.

Charles Tien is Professor of Political Science at Hunter College and the Graduate Center, CUNY. Michael S. Lewis-Beck is F. Wendell Miller Distinguished Professor of Political Science at the University of Iowa and co-author of the book Forecasting Elections.