Sylvia Kaufmann, "Dating and Forecasting Turning Points by Bayesian 
Clustering with Dynamic Structure: A Suggestion with an Application to 
Austrian Data", Journal of Applied Econometrics, Vo. 25, No. 2, 2010, 
pp. 309-344.

There are two data files, both ASCII files in DOS format, which are zipped
in the file kaufmann-data.zip. Unix users should use "unzip -a". There 
are also .xls versions of these two files, which are zipped in the file 
kauffman-xls.zip. Program files are zipped in the file kaufmann-progs.zip.

Data:

The files "datQ0703.txt" and "DAT_roh_Q0410.txt", and their .xls 
counterparts, are organized in the following way.

datQ0703.txt runs from 1988Q1 to 2006Q4
DAT_roh_Q0410.txt runs from 1988Q1 2004Q3 (estimation sample 1988Q1-2003Q4)

Observations are in rows and series are in columns. The first column
contains the dates. The first line contains the series labels (see table 1
in the paper, and below) and the second line contains an integer referring
to the classification of the series:

% 1   GDP and its main components
% 2   WIFO business cycle indicator
% 3   business surveys
% 4   prices (HICP)
% 5   CPI
% 6   wages
% 7   wholesale prices
% 8   foreign trade
% 9   labor market
% 10  IP
% 11  financial variables

The following series have been excluded for estimation (reason for exclusion):
'SZR' (inventory investment computed as statistical difference);
'YED';'PCD';'ITD';'GCD';'MTD';'XTD'; (all implicit deflators)
'ECOSEN' (older compilation of EECOS)
'INDSEN' (nearly perfectly correlated with EINDSE)
'ALQN' (not seasonally adjusted, nearly perfectly correlated with ALQNSA)

The following outliers have been identified and are replaced by a
missing-value estimate during model estimation.

datafile: datQ0703.txt
series:
'PCR' (1993Q1)
'ITR' (2003Q1)

datafile: DAT_roh_Q0410.txt
series:
'PCR' (1996Q2, 1996Q3)
'GCR' (1994Q4, 1995Q1, 1997)
'MTR' (1993Q1, 1993Q2)

Data:
YER    LD100  National account data GDP real
PCR    LD100  Private consumption, real  
ITR    LD100  Total investment, real
GCR    LD100  Government consumption, real
MTR    LD100  Imports, real  
XTR    LD100  Exports, real  
%SZR          Changes in inventories, real                      
%YED	      GDP deflator
%PCD	      Private consumption deflator
%ITD	      Gross fixed capital formation deflator
%GCD	      Public consumption deflator
%MTD	      Import eflator
%XTD          Export deflator
TOT    LD100  Terms of trade 
EEN    LD100  Nominal effective exchange rate on the export side
USD    LD100  Euro/US-Dollar exchange rate

       WIFO Quarterly survey

QTAUF      D  Assessment of order books, industry,total, balance in
QTEXPA     D  Assessment of export order books, industry, total, balance in 
QTLAG      D  Assessment of stocks, industry, total, balance in  
QTPR       D  Selling-price expectations, industry, total, balance in  
QTPRO      D  Production expectations, industry, total, balance in
QTBAUF     D  Assessment of order books, main construction trade, total,
              balance in   
QTBPR      D  Assessment of selling price expectations, main construction
              trade, total, balance in 
QTBBGL     D  Assessment of business situation, main construction trade,
              total, balance in   
QTBAGL     D  Evolution of business situation, main construction trade, total,
              balance in   
%ECOSEN       Economic sentiment indicator 
%INDSEN       Industry - Confidence indicator

       Monthly survey  data
       
EECOS  LD100  Economic sentiment indicator
EINDSE     D  Industry - Confidence indicator   
KTPROL     D  Industry - Production in recent months 
KTAUF      D  Industry - Order books total  
KTAUSL     D  Industry - Exports order books   
KTLAG      D  Industry - Stocks of finished goods 
KTPRON     D  Industry - Production over next months   
KTVPN      D  Industry - Selling prices over next months   
BAUVPN     D  Construction - Selling prices over next months
EBAUSE     D  Construction - Confidence indicator   
EHANSE     D  Retail trade - Confidence indicator
EKONSE     D  Consumer confidence indicator   
IFOERW LD100  Foreign confidence indicator   IFO - Business expectations
              in Western Germany, Index 1991 = 100
IFOKL  LD100  IFO - Business climate index for Western Germany, 1991=100
IFOGL  LD100  IFO - Business situation in Western Germany, Index 1991 = 100
PMI    LD100  Purchasing Manager Index USA  
HICP       LD100   HICP - Overall index   
HICP-FO    LD100   HICP - Food incl. alcohol and tobacco  
HICP-PF    LD100   HICP - Processed food incl. alcohol and tobacco
HICP-UF    LD100   HICP - Unprocessed food
HICP-G     LD100   HICP - Goods   
HICP-IG    LD100   HICP - Industrial goods
HICP-GX    LD100   HICP - Industrial goods excluding energy   
HICP-E     LD100   HICP - Energy  
HICP-S     LD100   HICP - Services
HICP-XA    LD100   HICP - All-items excluding alcoholic beverages, tobacco
HICP-XE    LD100   HICP - All-items excluding energy  
HICP-XF    LD100   HICP - All-items excluding energy and food 
HICP-XG    LD100   HICP - All-items excluding energy and unprocessed food 
VPIG86     LD100   CPI    Consumer price index, all items, 1986=100   
VPI-WOH86  LD100   CPI 86, housing, 1986=100 
TLIG86     LD100   Wages  Agreed minimum wages, overall index, 1986=100
TLIARG86   LD100   Agreed minimum wages, workers, index 1986=100  
TLIANG86   LD100   Agreed minimum wages, salary earners, index 1986=100   
GHPIG      LD100   Wholesale prices  - overall index, 1986=100   
GHPIOS     LD100   Wholesale prices  - excl. seasonal goods, 1986=100   
GHPIGG     LD100   Wholesale prices  - consumer goods, 1986=100    
GHPIGL     LD100   Wholesale prices  -  durable commodities,  1986=100 
GHPIGK     LD100   Wholesale prices  - non-durable commodities, 1986=100 
GHPIVBG    LD100   Wholesale prices  - non-durables, 1986=100  
GHPIKONG   LD100   Wholesale prices  - consumer goods, 1986=100 
GHPIINVG   LD100   Wholesale prices  - investment goods, 1986=100    
GHPIINTG   LD100   Wholesale prices  - intermediate goods, 1986=100
OEL        LD      Oil price, USD per Barrel - IMF 
EXPG       LD100   Exports - FOB total   
EXP6       LD100   Exports - FOB basic manufactures,  SITC 6   
EXP7       LD100   Exports - FOB machines and transport equipment, SITC 7 
EXP8       LD100   Exports - FOB misc. manufactured goods, SITC 8
IMPG       LD100   Imports - CIF total   
IMP6       LD100   Imports - CIF basic manufactures, SITC 6   
IMP7       LD100   Imports - CIF machines and transport equipment, SITC 7 
IMP8       LD100   Imports - CIF misc. manufactured goods, SITC 8
EXP-US     LD100   Exports of commodities to US 
EXP-EU     LD100   Exports of commodities to EU  
EXP-DE     LD100   Exports of commodities to Germany 
IMP-US     LD100   Imports of commodities from US  
IMP-EU     LD100   Imports of commodities from EU   
IMP-DE     LD100   Imports of commodities from Germany  
%ALQN              Unemployment rate, national definition, total 
ALQNSA     D       Unemployment rate, national definition, total SA
ALOSM      LD100   Unemployment, male  
ALOSW      LD100   Unemployment, female   
OFST       LD100   Vacancies   
STANDR     LD100   Employees 
INDPROD    LD100   IP Industrial production, overall index
                   (excl. construction and energy), 1995=100
ATX        LD100   ATX (Austrian Trading Index)
M1         LD100   M1 
M2         LD100   M2 
M3         LD100   M3 
DAX        LD100   DAX
DowJones   LD100   Dow Jones Index 
STI        D       3-month money market rate   
SEKMRE     D       Secondary market yield on government bonds (9 to 10 years)
YIELD      D       Yield spread   
DCR-H      LD100   Direct credits to private households 
DCR-F      LD100   Direct credits to private firms
DCR-G      LD100   Direct credits to government  
DEBT       LD100   Outstanding debt
DCR        LD100   Direct credits, total 


Programs:

The current settings allow to estimate the model with informative prior for
the data vintage ending in 2006: The program is implemented for the
following parametrization:

y_{it}= \mu^{i}_0-\mu^{i}_1*(1-I_{kt}) + \phi^{i}_1 * y_{i,t-1}+ \dots +
          \phi^{i}_p*y_{i,t-p}+\varepsilon_{it}

with \varepsilon_{it} ~ i.i.d N(0,\sigma^2/\lambda_i)
and I_{kt}=0 means contraction, I_{kt}=1 means expansion.

ESTIMATION of PDS (Panel with dynamic structure)

run 'start_model4_logit'

settings:
file='logit'; 
rest_diff=0; ieff_restr=0; post=0; 
not_switch=0; auto_switch=0; auto_grspec=1; 
pool=1; start_S=0; preclas=1; S_logit=1; DStruc=0; bas_cat=3; var_logit=[3];
eta_different=1; unit_spec_var=1; sg_change=1; 
dum_out=1; 
shrink=1;

true=0; span=18; maxlag=4; 
for cal_beg=[1988];
    cal_end=cal_beg+span;
    for cal_end_est=2006.75;
    for K=[3];
        for lag_dlo=[2]
        end
    end
    ....
end

output: MCMC output saved in 
logit1_dyn0_shr1_var1_perm0_sg1_K3end2_ex0_grspecstd_2006.75_2006


IDENTIFICATION, POSTERIOR RESULTS

run 'permut_MS_all'

settings:
span=18; pool=1; shrink=1; rest_diff=0; unit_spec_var=1; perm=0; sg_change=1; 
File='logit'; DSTruc=0; S_Logit=1; Pclas=1;
indexMSsort=[1 1 1];indexsort=1; ind_enc=[1 2]; 
for cal_beg=[1988];
    cal_end=cal_beg+span;
    for cal_end_Est=[2006.75];
    for K=3;
        for lag_dlo=[2]; 
            add=0; group=0; eta_restr=0;
.....
end

output: identified model and posterior moments saved in file
logit1_dyn0_shr1_var1_perm0_sg1_K3end2_ex0_grspecstd_2006.75_2006_iden  


GRAPH RESULTS

run 'plot_all'

span=18; rest_diff=0; pool=1; shrink=1; sg_change=1; unit_spec_var=1; perm=0;
File='logit'; DSTruc=0; S_Logit=1; Pclas=1;
for cal_beg=[1988];
    cal_end=cal_beg+span;
    for cal_end_Est=[2006.75];
    for K=[3];
        for lag_dlo=[2]; 
            lag_dir=0;
...
end

output: 
graphical evaluation of MCMC output:
graph 7 is figure 4
graph 9 is figure 6 in the paper
graphs 10,11 are in figure 3 of the paper
graphs 12,18 are in figure 2 
graph 16,17 are in figure 5 

TABULATED RESULTS

run 'output_rel_parms'
run 'output_ser_class'

note: run AFTER plot_all, DEFINE file='file.out' in the command line

output:
file in tex-format containing posterior moments of MCMC output, and series
classification

ESTIMATE PP,BDS and UNI

run 'start_model4_logit'

settings for PP (different from PDS):
file='pp'; S_logit=0; bas_cat=[]; var_logit=[]; for K=[1];

settings for BDS (different from PDS):
file='bds'; S_logit=0; bas_cat=[]; var_logit=[]; sg_change=0; shrink=0; for
K=[2];

settings for UNI (different from PDS):
file='uni'; S_logit=0; bas_cat=[]; var_logit=[]; unit_spec_var=0;
sg_change=0; shrink=0; for K=[1];

output: 
MCMC output saved in file pp...,bds...,uni...

ESTIMATE DYNAMIC STRUCTURE

run 'start_model4_logit'

settings (different from PDS):
preclas=0; S_logit=0; DStruc=1; bas_cat=[]; var_logit=[];

output:
MCMC output saved in file 
logit0_pclas0_dyn1_shr1_var1_perm0_sg1_K3end2_ex0_grspecstd_2006.75_2006

ESTIMATE PDS, PP,BDS,UNI for out-of-sample forecasting

run 'start_model4_logit'

settings: see above
ADDITIONAL setting:
 for cal_end_est=[2000.00:0.25:2006.75];

output:
MCMC output saved in files logit..., pp...,bds...,uni...

IN-SAMPLE FORECASTING

needs the file 'dating.mat' in 'plot_forecast' containing the posterior
estimates P(I_t=1|Y^T)>0.5 for the coincident group obtained from the
various models: ecri, pds, pp, bds, uni (and also itl)
 
run 'start_forecast'

For PDS:
settings:
span=18; pool=1; Shrink=1; Unit_spec_var=1; sg_change=1;
File='logit'; DSTruc=0; S_Logit=1; rest_diff=0; perm=0; 
for cal_beg=[1988];
    cal_end=cal_beg+span;
    for cal_end_est=[2006.75];
    for K=[3];
        for lag_dlo=[2];
...
end

For PP:
settings (different from PDS):
File='pp'; S_Logit=0; K=[1];

For BDS:
settings (different from PDS):
Shrink=0; Unit_spec_var=1; sg_change=0;
File='bds'; S_Logit=0; 
K=[2];

For UNI
settings(different from PDS):
Shrink=0; Unit_spec_var=0; sg_change=0;
File='uni'; S_Logit=0; K=[1];

output:
graphs summarizing the forecasts
file forecast.out containing model-based evaluation and evaluation against
ecri and pds

OUT-OF-SAMPLE FORECASTING

needs the file 'dating.mat' in 'plot_forecast' containing the posterior
estimates P(I_t=1|Y^T)>0.5 for the coincident group obtained for the various
models evaluated: ecri, pds, pp, bds, uni (and also itl)

run 'start_forecast_outsample'

For PDS:
settings:
file='logit'; span=18; Pool=1; Shrink=1; Sg_change=1; Unit_spec_var=1; perm=0;
ifig=7;
for cal_beg=[1988];
    cal_end=cal_beg+span;
    for cal_end_EST=[2000:0.25:2006.75];
    for KK=[3];
...
end

For PP:
settings (different from PDS):
file='pp'; for KK=[1];

For BDS:
settings (different from PDS):
file='bds'; Shrink=0; Sg_change=0; KK=[2];

For UNI:
settings (different from PDS):
file='uni'; Shrink=0; Sg_change=0; Unit_spec_var=0;
KK=[1];

output:
graphs summarizing the forecasts
file forecast_outsample.out containing model-based evaluation and evaluation
against ecri and pds

EVALUATE SCENARIOS

needs the file 'scenarios.mat' containing:
si: the posterior probability of contraction 
ci: the posterior classification probability into the coincident group
li: the posterior classification probability into the leading group

estimated under the various scenarios:
(i) \rho^* fixed, logit prior, full pre-classification
(ii) \rho^* estimated, no pre-classification
(iii) \rho^* fixed, full pre-classification
(iv) \rho^* fixed, no pre-classification
(v) \rho^* fixed, logit prior, partial pre-classification
(vi) \rho^* fixed, partial pre-classification
(vii) \rho^* fixed, full pre-classification, grouping state indicator only
(viii) \rho^* fixed, partial pre-classification, grouping state indicator only

the scenarios are estimated by running 'start_model4_logit' with the
following settings:

(i) PDS settings, see above; for all scenarios: to use data vintage ending
    in 2003 set span=15, cal_end_est=2003.75;

(ii) settings (different from PDS): 
     preclas=0; S_logit=0; DStruc=1; bas_cat=[]; var_logit=[]; sg_change=1; 

(iii) settings (different from PDS);
      S_logit=0; bas_cat=[]; var_logit=[];

(iv) settings (different from PDS);
     preclas=0; S_logit=0; bas_cat=[]; var_logit=[];

(v) PDS settings, see above; 
    edit daten_read_cont0703.m (daten_read_cont0410.m), lines 28 and 29:
    ser_cont={'YER'};%'PCR';'ITR';'MTR';'XTR'}; %series falling into the
      contemporaneous group
    ser_lead={'KTAUF'};%'KTPROL';'QTAUF';'EINDSE'}; %series falling into the
      leading group

(vi) settings (different from PDS);
     S_logit=0; bas_cat=[]; var_logit=[];
     edit daten_read_cont0703.m (daten_read_cont0410.m), lines 28 and 29:
     ser_cont={'YER'};%'PCR';'ITR';'MTR';'XTR'}; %series falling into the
       contemporaneous group    
     ser_lead={'KTAUF'};%'KTPROL';'QTAUF';'EINDSE'}; %series falling into
       the leading group

(vii), (viii), the programs are available upon request

run 'eval_scenarios'

OTHER FIGURES

figure 1 in the paper
run 'data_properties'

output: graph 2-4 are in figure 1 of the paper
data_prop.tex contains tables in tex-format summarizing data properties
saves file corr_0703.asc containing the correlations with GDP and KTAUF
shaping the informative prior on classification probability

figure 7
run 'plot_dat'