R package which implements estimation of SVARs using sign restrictions. Also includes narrative restrictions as discussed in AR18.

To install run:

devtools::install_github("aaron-lohmann/narrSVAR")

To work through a minimal example with stored data. Technical details, refer to vignitte or AR18.

# Application of Uhlig (2005) - Monetary policy shock

# load data
data("uhlig")
# set up traditional sign restrictions
shock_names_uhlig <- c("Monetary Policy")

restr_matrix_uhlig <- matrix(NA, 6, 6)
restr_matrix_uhlig[1, ] <- c(NA, -2, -3, NA, -5, 6)
hor_matrix_uhlig <- matrix(NA, 6, 2)
hor_matrix_uhlig[1, ] <- c(1, 6)

uhlig_tradsign <- tradsign_setup(
  shock_names = shock_names_uhlig,
  restr_matrix = restr_matrix_uhlig,
  hor_matrix = hor_matrix_uhlig,
  cum = rep(0, 6)
)
# set up narrative sign restrictions
# uhlig
varnames_uhlig <- colnames(uhlig)
allshocknames <- c("MP")
shock_names <- c("MP", "MP")
shock_type <- c("sign", "contribution")
shock_dates <-
  matrix(c("1979-10-01", "1979-10-01", "1979-10-01", "1979-10-01"), 2, 2)
dates <- uhlig[, 1]
relevant_variable <- c(NA, varnames_uhlig[7])
# need to explain how to select shock_sign variable
shock_sign <- c(1, 1)
# strong or weak
shock_size <- c(NA, "strong")
lag <- 12
data <- uhlig[, -1]

uhlig_narr_rest <- narrsign_setup(
  allshocknames = allshocknames,
  shock_names = shock_names,
  shock_type = shock_type,
  shock_dates = shock_dates,
  dates = dates,
  relevant_variable = relevant_variable,
  shock_sign = shock_sign,
  shock_size = shock_size,
  lag = lag,
  data = data
)

# estimate the model with traditional and narrative sign restrictions
uhlig_m <- narrsign(
  data = ts(uhlig[, -1]),
  lags = 12,
  trad_signs = uhlig_tradsign,
  narr_restr = uhlig_narr_rest,
  nkeep = 1000,
  draws = 100,
  subdraws = 300,
  elasticitybounds = NULL,
  const = FALSE,
  steps = 30
)

# plot the irfs for monetary policy shock. Notice the scaling component

scaling_trad <- 0.25 / median(uhlig_m$trad$IRFS[, 1, 1, 6], na.rm = TRUE)
scaling_narr <- 0.25 / median(uhlig_m$narr$IRFS[, 1, 1, 6], na.rm = TRUE)

two_irf_plot(
  uhlig_m,
  whichShock = 1,
  scaling_trad = scaling_trad,
  scaling_nar = scaling_narr,
  varnames = c(
    "Output",
    "GDP Deflator",
    "Commodity Prices",
    "Total Reserves",
    "Non-Borrowed Reserves",
    "Federal Funds"
  )
)