State - dependant local projections model

[Adriano Manferdini]

I am running a Local Projections model to try and estimate how household net worth affects the effectiveness of monetary policy. My code for the the LP model is as follows:

locproj log_real_gdp, hor(12) shock(l.fed_rate) controls(l(1/4).log_hh_networth l(1/4).log_gdp_deflator l(1/4).spread l(1/4).log_exchange_rate) met(newey) hopt(lag) ylags(4)

The dependent variable is log_real_gdp and the shock is the fed funds rate lagged one period back. The controls all contain 4 lags.

I want to interact this model with a dummy variable that takes on a value of 1 for low household net worth cycles and 0 for high household net worth cycles.

I generated my dummy variable as follows:

tsfilter hp m = log_hh_networth, smooth(1600) trend(networth_growth)

gen low_hhnw = 0

replace low_hhnw = 1 if m < 0

How would I then interact the lagged dummy variable with the model above so I can uncover monetary policy effectiveness depending on the household net worth state?

[Alfonso Ugarte-Ruiz]

Hi, I don't know if you have already solved your problem. I think I would write the following: I'm going to treat your dummy variable as a continuous variable and I'm going to introduce the fed_rate and the interaction with the dummy (fed_rate + c.low_hhnw#c.fed_rate) with a lag:


1. If you want the IRF of log_real_gdp to a shock in l.fed_rate when low_hhnw=1

locproj log_real_gdp, hor(12) shock(l.(fed_rate + c.low_hhnw#c.fed_rate)) controls(l(1/4).log_hh_networth l(1/4).log_gdp_deflator l(1/4).spread l(1/4).log_exchange_rate) met(newey) hopt(lag) ylags(4)

2. If you want the IRF of log_real_gdp to a shock in l.fed_rate when low_hhnw=0

locproj log_real_gdp, hor(12) shock(l.fed_rate) controls(c.l.low_hhnw#c.l.fed_rate l(1/4).log_hh_networth l(1/4).log_gdp_deflator l(1/4).spread l(1/4).log_exchange_rate) met(newey) hopt(lag) ylags(4)