

1. Optimization and MCMC
The heart of the estimation is the optimization of the posterior likelihood and then the MCMC. The optimization is run by optimize14_bonds.m. I ran this a few hundred times in parallel since it often gets stuck in local extrema. that optimization file requires Chris Sims's csminwel algorithm, which is available on his website (http://sims.princeton.edu/yftp/optimize/). 

The markov chain is run by MCMC_slow.m. 

Table 1: table_1.m

Table 2: first two rows are from tabs-figs.m. The remaining rows are from the model at the posterior mode. 

Table 3: main_full14_bonds

Table 4: run IRF_posterior.m. Then use the outputs VVD_025, VVD_500, and VVD_975 to get the confidence intervals for the bar chart in the bottom. The variance decomposition itself is the output VD1'.

Table 5: tabs-figs.m

Table 6: print-params.m (but using the alternative posterior maxima).

Figure 1: The raw data is contained in the spreadsheet it requires no other code.

Figure 2: tabs-figs.m

Figure 3: tabs-figs.m

Figure 4: tabs-figs.m

Figure 5: tabs-figs.m (use dLEV, dSLO, and dCUR to get the data that is in the spreadsheet)

Figure 6: tabs-figs.m

Figure 7: IRFs are created in main_full14_bonds.m

Figure 8: tabs-figs.m

Figure 9: variance decomposition.m







