'Author: Nwabisa Florence Ndzama 'Corvinus Univeristy of Budapest !read = 1 !Nwabisa =1'0 !Zsolt = 0 '1 !stop_after_read = 0 if !read = 1 then '---------------------------------------------------------------------------------- close CA_forecast_sustainability_probabilities.wf1 wfcreate(wf=CA_forecast_sustainability_probabilities, page = Read) a 1948 2030 smpl 1948 2029 'Actual data starts from 1960. The "period" 1948 -1959 are data decsription colums from the Excel spreadsheet. 1955=0, its the scale (one of descriptive items.) For %sheet CA GDP_growth Trade_Openness REER_ZD Budget_balance OAD if !Nwabisa = 1 then read(c3,s={%sheet},t)"File Path" 204 ' 204 is the number of observations (countries + regions in the sample) from excel file endif if !Zsolt = 1 then ' read(c3,s={%sheet},t) "C:\Users\Zsolt\Dropbox\2023-04 CA reversals\Data\WEO_data_for_probit.xls" 204 read(c3,s={%sheet},t)"File Path" 204 ' 204 is the number of observations (countries + regions in the sample) from excel file endif Next ' %sheet 'stop ' these series are not included in the dataset; we create them (with NA values) to eliminate some error messages when this code is running series REER_ZD_SOM series REER_ZD_TLS series REER_ZD_PRI '1948 value in the excel file is the numerical code of the country. To avoid confusion of these numerical codes with actual data, we create a new sheet for the sample 1980-2027 and copy all data there: Pagecreate(page=Annual) a 1960 2029 '1977- 2028 , Starting few years before our year (1983) of interest (due to data for all other variables in the analysis) copy Read\* Annual\* if !stop_after_read = 1 then Statusline Stopped after reading the data STOP endif endif !read=1 'STOP group a_group_BUDGET_BALANCE BUDGET_BALANCE* 'show a_group_BUDGET_BALANCE group a_group_ca ca* 'show a_group_ca group a_group_GDP_growth GDP_growth* 'show a_group_GDP_growth group a_group_Trade_Openness Trade_Openness* 'show a_group_Trade_Openness group a_group_OAD OAD* 'show a_group_OAD group a_group_REER_ZD REER_ZD* 'show a_group_REER_ZD 'stop 'Pagecreate(page=distributionVAR) 1000000 !forc_period_ahead =5' 1 2 3 4 5 number of periods ahead for which out-of-sample forecasts will be made from each date. For Annual DATA, 5 means 5 years. !n = 1000000 '1000 ' The number of observations (e.g. the number of forecasts made) !threshold = -8 '-8 '-5 '-4 '-2 ' The threshold value: this code calculates the probability that an observation is BELOW this value. '================================================================================================= 'Nwabisa 22/06/2024 MODEL TYPE %forecast_model = "STVAR" ' AR, VAR, BVAR,vecm, BTVC, STVAR, SWTVAR '==================================================================================================== 'Pageselect Annual table table_countries '!i=0 !c=0 For %c AUS 'FRA GRC IRL ESP 'BGR CYP HUN ITA PRT ROU 'FRA GRC IRL PRT ROU ESP 'AUS 'CHL KOR MEX USA 'ZAF 'HUN MYS' GRC 'LVA ISR HUN USA 'DEU 'IND IDN MEX HUN RUS ZAF AUS GBR FRA DEU USA 'CHN 'Prelimenary sample, Chapter 5 'For %c GRC 'HUN GRC 'IRL MYS ZAF 'HUN 'MYS 'HUN ' for this sample the model excludes the budget balance variable due to data unavailability , vcem model : we exclude TO and BB for Hungary due to Data limitations (error message: insufficient parameters) 'For %c ALB ARG ARM AUS AUT AZE BGD BLR BEL BGR BOL BRA BRN BDI CPV KHM CMR CAN CAF TCD CHL COL CRI CIV CYP DNK DOM DZA ECU EGY SLV FJI FIN FRA GAB DEU GHA GRC GTM GIN HTI HND HKG HUN ISL IND IDN IRL ITA KEN KOR MDG MYS MLI MUS MEX MNG MAR NAM NPL NLD NIC NER MKD NOR PAK PRY PER PHL PRT ROU RUS RWA SAU SEN SYC SGP ZAF ESP SDN SWE CHE TZA THA BHS GMB TGO TUN TUR UGA UKR GBR USA 'Extended SAMPLE, CHAPTER 5 'BGR JAM JOR KGZ LKA 'UPDATED FULL/EXTENDED SAMPLE - 8/10/2024 ' For %c ALB DZA ARG ARM AUS AUT AZE BGD BLR BEL BLZ BOL BRA BRN BGR BDI CPV KHM CMR CAN CAF TCD CHL COL CRI CIV CYP DNK DOM ECU EGY SLV FJI FIN FRA GAB DEU GHA GRC GTM GIN HTI HND HKG HUN ISL IND IDN IRL ITA JAM JOR KEN KOR KGZ MDG MYS MLI MUS MEX MNG MAR NAM NPL NLD NIC NER MKD NOR PAK PRY PER PHL PRT ROU RUS RWA SAU SEN SYC SGP ZAF ESP LKA SDN SWE CHE TZA THA BHS GMB TGO TUN TUR UGA UKR GBR USA 'For %c BWA 'NAM SWZ ZAF 'BWA LSO BGR USA 'CYP FRA GRC HUN IRL ITA PRT ROU ESP 'BGR ' !i=!i+1 !c=!c+1 Pageselect Annual table_countries(!c,1) = %c 'Next '%c! %start_est_date = "1980" ' the first date for the estimation of models %start_forecast_date = "2005" '2005 is the first date for which a forecast will be made, therefore the recursive estimation goes upto 2004, that is 1980-2004 %last_Y_obs = "2023" ' the last observation of the sample period !forc_period_ahead = 5' 1 2 3 4 5 number of periods for which out-of-sample forecasts will be made. if its Annual DATA, 5 means 5 years. !start_forecast_date = @val(%start_forecast_date) !last_Y_obs = @val(%last_Y_obs) !num_forecasts = !last_Y_obs - !start_forecast_date ' e.g. if the %start_forecast_date = "2017" and %last_Y_obs = "2018" , then !num_forecasts=1 ' !num_forecasts = !last_Y_obs - !start_forecast_date +1 '------------------------------------------------------------------------------------------------------------------------------------------------------------ 'SMOOTH TRANSITION VAR MODEL (ST-VAR) ' Step 1: Estimate the Model 'ADDED ON 09/10/2024 'Strength: Unlike traditional linear VAR models, ST-VAR allows for nonlinear dynamics and captures abprut changes in a times series. '------------------------------------------------------------------------------------------------------------------------------------------------------------ if %forecast_model = "STVAR" then Pageselect Annual table table_probabilitiesSTVAR 'Create a table that collects the probability results table_probabilitiesSTVAR(1,1) = "This table shows the probability that the CA balance will be lower than a threshold value" table_probabilitiesSTVAR(2,1) = "model: " table_probabilitiesSTVAR(2,2) = "STVAR" table_probabilitiesSTVAR(3,1) = "number of forecasts (!n): " table_probabilitiesSTVAR(3,2) = !n table_probabilitiesSTVAR(4,1) = "threshold: " table_probabilitiesSTVAR(4,2) = !threshold table_probabilitiesSTVAR(5,1) = "Forecast Period: " table_probabilitiesSTVAR(5,2) = "1 year ahead" table_probabilitiesSTVAR(5,3) = "2 year ahead" table_probabilitiesSTVAR(5,4) = "3 year ahead" table_probabilitiesSTVAR(5,5) = "4 year ahead" table_probabilitiesSTVAR(5,6) = "5 year ahead" table_probabilitiesSTVAR(6,1) = "country" table_probabilitiesSTVAR(6,2) = "probability" table_probabilitiesSTVAR(6,3) = "probability" table_probabilitiesSTVAR(6,4) = "probability" table_probabilitiesSTVAR(6,5) = "probability" table_probabilitiesSTVAR(6,6) = "probability" ' Zsolt 01/06/2024: important to add this info to the table, so you'll surely know for which period the analysis is made table_probabilitiesSTVAR(5,8) = "%start_est_date" table_probabilitiesSTVAR(5,9) = "%start_forecast_date" table_probabilitiesSTVAR(5,10) = "%last_Y_obs" ' 24 Aug 2024: I added the CA balance value at %last_Y_obs, and the forecasts for the next 5 years, which help to the check visually if estimated probabilities look correct or not table_probabilitiesSTVAR(4,12) = "Actual CA balance" table_probabilitiesSTVAR(5,12) = "%last_Y_obs" table_probabilitiesSTVAR(4,13) = "Forecast CA balance" table_probabilitiesSTVAR(5,13) = "%last_Y_obs+1" table_probabilitiesSTVAR(4,14) = "Forecast CA balance" table_probabilitiesSTVAR(5,14) = "%last_Y_obs+2" table_probabilitiesSTVAR(4,15) = "Forecast CA balance" table_probabilitiesSTVAR(5,15) = "%last_Y_obs+3" table_probabilitiesSTVAR(4,16) = "Forecast CA balance" table_probabilitiesSTVAR(5,16) = "%last_Y_obs+4" table_probabilitiesSTVAR(4,17) = "Forecast CA balance" table_probabilitiesSTVAR(5,17) = "%last_Y_obs+5" ' Zsolt 01/06/2024: for each country, we create a separate page, e.g. for South Africa the name of the page will be: distributionVAR_ZAF %pagename = "distributionSTVAR" + "_" + %c ' Zsolt 01/06/2024: we add this info to the table, so you'll surely know for which period the analysis is made table_probabilitiesSTVAR(6+!c,1) = %c table_probabilitiesSTVAR(6+!c,8) = %start_est_date table_probabilitiesSTVAR(6+!c,9) = %start_forecast_date table_probabilitiesSTVAR(6+!c,10) = %last_Y_obs Pagecreate(page={%pagename}) U !n Pageselect Annual 'var estimates For !smpl = 0 to !num_forecasts smpl %start_est_date %start_forecast_date-1+!smpl ' so when %start_forecast_date = "2017" and !smpl = 0, you estimate the model till 2016 and make a forecast for 2017 'var ca_{%c}_var06.threhold(method=fixedseq, type=smooth) 1 1 CA_{%c} GDP_GROWTH_{%c} REER_ZD_{%c} OAD_{%c} TRADE_OPENNESS_{%c} BUDGET_BALANCE_{%c} @thresh 6 'equation ca_{%c}_var06.threhold(method=fixedseq, type=smooth) CA_{%c} GDP_GROWTH_{%c} REER_ZD_{%c} OAD_{%c} TRADE_OPENNESS_{%c} BUDGET_BALANCE_{%c} @thresh 6 '' NOTE: I changed the lag lenght to 2 so that fewer parameters have to be estimated, which allows a longer out-of-sample forecast period 'equation eq1.threshold(method=fixedseq, type=discrete) ss_transf c ss_transf(-1 to -11) @thresh 2 'show ca_{%c}_var06 '================CREATING A SYSTEM OF EQUATIONS TO STVAR=========================================== system STVAR STVAR.append CA_{%c} = c(1)+c(2)*CA_{%c}(-1)+c(3)*GDP_GROWTH_{%c}(-1)+(1-EXP( -c(4)*(CA_{%c}(-1)-c(5))^2 ) )*(c(6)+c(7)*CA_{%c}(-1)+c(8)*GDP_GROWTH_{%c}(-1)) STVAR.append GDP_GROWTH_{%c} = C(9)+C(10)*GDP_GROWTH_{%c}(-1)+c(11)*CA_{%c}(-1)+(1-EXP( -c(4)*(GDP_GROWTH_{%c}(-1)-c(5))^2 ))*(c(12)+c(13)*GDP_GROWTH_{%c}(-1)+c(14)*CA_{%c}(-1)) STVAR.append CA_{%c}(-1) GDP_GROWTH_{%c}(-1) STVAR.nls ' estimate the parameters of a model using nonlinear least squares. show STVAR.results Next !smpl endif '%forecast_model = "STVAR" then Next '!c Stop