sectoral slowdowns in the uk: evidence from transmission probabilities and economic linkages (replication files)

This folder contains MATLAB and R software and data to accompany the paper "Sectoral slowdowns in the UK: Evidence from transmission probabilities and economic linkages" by E.F. Janssens and R.L. Lumsdaine published in the Journal of Applied Econometrics.

This version: Feb 2022. The following items are provided:

1. LICENSE AGREEMENT (CC BY-NC-SA 4.0): The software is distributed under a Creative Commons Attribution NonCommericial-ShareAlike (CC BY-NC-SA) 4.0 International Public License, available at https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode .
   The license agreement explains the terms and conditions under which you can use, share and adapt this Software.

2. DATA:

We have directly copied the original data into the Matlab workspace file 'workspace_2020_07_01.mat'. You don't have to do anything with this, the relevant scripts you run will load this data into Matlab for you. The original data sources are:

For the whom-to-whom matrices: https://www.ons.gov.uk/economy/nationalaccounts/uksectoraccounts/datasets/enhancedfinancialaccountsflowoffundstotalfinancialaccountsexperimentalstatistics Here we use what they label as 'the 2018 edition'.

For the total financial assets, liabilities and net worth https://www.ons.gov.uk/economy/nationalaccounts/uksectoraccounts/datasets/unitedkingdomeconomicaccountsflowoffunds/current Here we use the data released on the 31th of March 2020.

These data are made available under the Open Government License v3.0, so we are allowed to redistribute them.

We provide you these data in csv files, under the names 'enhanced_financialaccounts.csv' and 'financial_assets.csv', 'financial_liabilities.csv' and 'financial_networth.csv'. These are the same data as in 'workspace_2020_07_01.mat', which we elaborate on below:

'workspace_2020_07_01.mat' contains the following variables:

financialaccounts_matrix (as in 'enhanced_financialaccounts.csv'): this matrix is 144x90, which is because the initial data comes with 11 sectors (+unknown) and 90 time observations, and this matrix essentially stacks the financial flow matrices between all sectors. Later in our files we merge certain sectors and will end up with a total of 8 sectors (note it also lists NMMF which is a subsection of OFI so we get rid of that too, so there are actually only 10 sectors in this set already). These data are obtained from the enhanced financial accounts linked above. Goes from Jan 1997 to April 2019.

Flowoffundsbasedassets (as in 'financial_assets.csv'), Flowoffundsbasedliab (as in 'financial_liabilities.csv'), Flowoffundsbasednetworth (as in 'financial_networth.csv'): these matrices contain the total assets, total liabilities and total net worth obtained from the flow of funds data linked above. These have 10 different sectors and a longer time period (133 periods, note that the last values in the matrices are nan's which is why their length does not correspond to 133): goes from Jan 1987 to Jan 2020.

N: number of sectors in the financial accounts matrix = 12, but after merging and cleaning this will go to 8.

Sectors: string 1x12 containing all the sector names of the enhanced financial accounts data

T: number of time observations of the enhanced financial accounts matrix, equals 90

Time: months and year for each data observation in the enhanced financial accounts data (1x90 datetime datatype)

1. CODE:

This folder has three scripts you need to run to replicate the results in the paper. The code does not work if you deviate from this order. That is, you need to first run (i) before (ii) and (iii).

(i) NetworkAnalysis_new_fof.m:

This script generates the maximum likelihood estimates. Run this script.

Useful output: - p_i_mode_mode_alt_stdev: this matrix contains the mode and standard deviation of the MLE of p_i, corresponding to those in Table 1 of the paper. Use the second and third row of the matrix. Also in Table 3 and Table E1. - q_i_mode_mode_alt_stdev: this matrix contains the mode and standard deviation of the MLE of p_i, corresponding to those in Table 1 of the paper. Use the second and third row of the matrix. Also in Table 3 and Table E1. - R0c_mean_modealt_stdev: these statistics are reported in Table 1 - R0d_mean_modealt_stdev: these statistics are reported in Table 1 - pijs_mode_mle_alt: these are the values of the pijs reported in Table 1 - pijs_stdev_mle: these are the standard errors of the pijs reported in Table 1

Apart from these statistics, this script generates several figures given in the paper:

Figure C4 in the Online Appendix is figure 11 in the script. Figure C1 in the Online Appendix is made by figure 6 and 7 in the script. Figure C3 in the Online Appendix is made by figure 101 in the script.

Figure D1 in the Online Appendix is made by figure 23 in the script. Figure F3 in the Online Appendix is made by figure 12 and 13 in the script.

(ii) NetworkAnalysis_new_withpriors_fof.m:

This script generates the Bayesian estimates with scaling factor. Run this script.

Useful output: - pijs_mode_alt: reported in Table 3 - pijs_se: reported in Table 3 - prior_pijs_mean: reported in Table 2 - prior_pijs_se: reported in Table 2 - R0c_mean_mode_var: reported in Table 3 - R0d_mean_mode_var: reported in Table 3 - z_mean_mode_var: reported in Table 3 - n_mean_mode_var: reported in text

In addition, the script produces several figures.

Figure 3 of the main paper is figure 18 in the script. Figure F1 is figure 23 in the script.

(iii) NetworkAnalysis_new_withpriors_noscale_fof.m: This script generates the Bayesian estimates without the scaling factor (z=1). Run this script.

Useful output - pijs_mode_alt: reported in Table E1 - pijs_se: reported in Table E1 - R0c_mean_mode_var: reported in Table E1 - R0d_mean_mode_var: reported in Table E1

In addition, the script generates among others the following figure: Figure E1 is figure 23 in the script