Run a simple bi-directional Granger Causality model of change in residential investment and GDP. It turns out a model with one and three lags best fits the data going back to 1969. That model's four quarter forecast for GDP growth is 2.6, 2.5, 2.1 and 2.5 percent; for residential investment growth is 6.2, 5.0, 4.9 and 4.9 percent. (BTW, the model passes the stationarity test).
But residential investment has grown by between 8.5 and 20 percent over the past four quarters. Let's say that an exogenous shock (kids moving out of their parents' houses) leads residential investment to grow by 10 percent. The forecast for GDP growth now increases to 2.6, 2.9, 2.6 and 2.9, or about .4 percentage points higher than the baseline case. This increase in GDP reflects more than the direct impact of residential investment on GDP.