André K. Anundsen, Karsten Gerdrup, Frank Hansen, and Kasper Kragh-Sřrensen, "Bubbles and Crises: The Role of House Prices and Credit", Journal of Applied Econometrics, Vol. 31, No. 7, 2016, pp. 1291-1311. The data used in the article were collected from various sources and cover the period from 1975q1-2014q4 for 20 OECD countries. The data set contains the following countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Italy, Japan, Korea, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, UK, USA. The data are provided both in .dta format, for the calculations in STATA, and in .csv format, and is organized as a panel. In addition, for country-specific unit root tests (where Ox is used), data are provided in an .xls file, where the different sheets provide the time series data for each country. Because some files are binary, Unix/Linux users should *not* use unzip with the "-a" option. ASCII files in DOS format will need to be converted to Unix format individually. Note: Some variables are not available at all horizons for all countries, and not all variables are available for all countries. Hence, it is an unbalanced panel. The following data series are included: Identifiers: - Country: String variable with country abbreviation - country_id: Numeric variable to identify the different countries - quarter: Lists the date (quarter) of the observation - year: The year of observation - period: Identifier to keep track on obervation number for each country, i.e. year-quarter observations GDP data: - nGDP: Nominal GDP, Source: OECD - rGDP: Real GDP, Source: OECD - gdp_gap: HP filtered GDP gap (lambda = 3000), Source: OECD CPI data: - CPI: Consumer price index, Source: OECD Household credit data: - hcredit: Nominal household credit, Source: BIS - real_hcredit: Real household credit, Source: BIS and OECD - hcredit_gdp: Household credit relative to GDP, Source: BIS and OECD - hcredit_gdp_gap: HP filtered gap of household credit relative to GDP (lambda = 400,000), Source: BIS and OECD Non-financial enterprise credit data: - nfecredit: Nominal non-financial enterprise credit, Source: BIS and OECD - real_nfecredit: Real non-financial enterprise credit, Source: BIS - nfecredit_gdp: Non-financial enterprise credit relative to GDP, Source: BIS and OECD - nfecredit_gdp_gap: HP filtered gap of non-financial enterprise credit relative to GDP (lambda = 400,000), Source: BIS and OECD Total private sector credit data: - pcredit: Nominal private sector credit, Source: BIS - real_pcredit: Real private sector credit, Source: BIS and OECD - pcredit_gdp: Private sector credit relative to GDP, Source: BIS and OECD - pcredit_gdp_gap: HP filtered gap of private sector credit relative to GDP (lambda = 400,000), Source: BIS and OECD Banking data: - wholesale_assets: Non-core funding relative to total assets, Source: OECD Banking Statistics - wholesale_gap: HP filtered gap of non-core funding relative to total assets (lambda = 400,000), Source: OECD Banking Statistics - equity_ratio: Capital and reserves in banking sector relative to total assets, Source: OECD Banking Statistics House price data: - real_hp: Real housing prices, Source: International house price database of Dallas Fed, as well as OECD - hp_disp_inc: Housing prices relative to disposable income, Source: International house price database of Dallas Fed, as well as OECD - hp_inc_gap: HP filtered gap of housing prices relative to disposable income (lamda = 400,000), Source: International house price database of Dallas Fed, as well as OECD Exuberance data: All exuberance measures are calculated using tests for explosive roots (see Phillips et al. (2011), Phillips et al. (2015b), and Phillips et al. (2015a)). They are constructed by subtracting 5% simulated critical values from the test statistics. All codes for the underlying calculations are available upon request. MatLab codes to calculate exuberance data are available at the personal homepage of Shu-Ping Shi: https://sites.google.com/site/shupingshi/home/research - exub_hp_inc: Exuberance in house prices relative to income, Source: International house price database of Dallas Fed, as well as OECD - ind_hp_inc_exu: Indicator variable taking the value 1 if "exub_hp_inc">=0 and 0 otherwise - exub_cred_g: Exuberance in credit relative to GDP, Source: International house price database of Dallas Fed, as well as OECD - ind_credit_gdp_exu: Indicator variable taking the value 1 if "exub_cred_g">=0 and 0 otherwise Global data using trade weights: The baseline global data are calculated using time-varying trade weights (see Pesaran et al. (2004), Dees et al. (2007a) and Dees et al. (2007b)) The GVAR data can be downloaded freely from: https://sites.google.com/site/gvarmodelling/gvar-toolbox - global_real_credit: Global real credit. Constructed by weighting "real_pcredit" using GVAR trade weights, Source: BIS - global_credit_gdp: Global credit relative to GDP. Constructed by weighting "pcredit_gdp" using GVAR trade weights, Source: BIS - global_credit_gdp_gap: Global credit to GDP gap. Constructed by weighting "pcredit_gdp_gap" using GVAR trade weights, Source: BIS - global_real_hp: Global real house prices. Constructed by weighting "real_hp" using GVAR trade weights, Source: BIS - global_hp_inc: Global house prices relative to GDP. Constructed by weighting "hp_disp_inc" using GVAR trade weights, Source: BIS - global_hp_inc_gap: Gap in global house prices relative to income. Constructed by weighting "hp_inc_gap" using GVAR trade weights, Source: BIS Global data using equal weights: - global_real_credit_equal: Global real credit. Constructed by weighting "real_pcredit" using equal weights - global_credit_gdp_equal: Global credit relative to GDP. Constructed by weighring "pcredit_gdp" using equal weights, Source: BIS - global_credit_gdp_gap_equal: Global credit to GDP gap. Constructed by weighting "pcredit_gdp_gap" using equal weights, Source: BIS - global_real_hp_equal: Global real house prices. Constructed by weighting "real_hp" using equal weights, Source: Dallas Fed, as well as OECD - global_hp_inc_equal: Global house prices relative to GDP. Constructed by weighting "hp_disp_inc" using equal weights, Source: Dallas Fed, as well as OECD - global_hp_inc_gap_equal: Gap in global house prices relative to income. Constructed by weighting "hp_inc_gap" using equal weights, Source: Dallas Fed, as well as OECD The file aghk-files.zip contains all the files organized under a folder called "Replication". It is organized as two main folders named "Paper" and "Appendix". They are organized as follows: 1. "Paper": Files needed to replicate all results in the paper. The following subfolders are contained in it: a) "Data": The data needed to replicate the results in the paper are provided in both .csv and .dta format b) "Do files": All the STATA do-files needed to replicate results in the paper. Edit line 15 in "main_paper.do" to change the directory to the place you save the folder. After this, the program can be executed part-by-part, or in one go, without any further changes to the program. The name of the sub-routines (the do files called my "main_paper.do") are consistent with the numbering of figures and tables in the paper. c) "Figures": This is where all figures reported in the paper are saved in .pdf format (empty until the program is executed) d) "Output": Measures of relative usefulness, true positive rates, false positive rates are saved in a .csv file in this folder (empty until the program is executed) e) "Tables": LaTex tables reported in the paper that summarize the results reported in the paper (empty until the program is executed) 2. "Appendix": Files needed to replicate all results in the online appendix. The following subfolders are contained: a) "Data": The data needed to replicate results reported in the online appendix are provided in both .csv and .dta format b) "Do files": All the STATA do-files needed to replicate results in the online appendix. Edit line 15 in "main_appendix.do" to change the directory to the place you save the folder. After this, the program can be executed part-by-part, or in one go, without any further changes to the program. The name of the sub-routines (the do files called my "main_appendix.do") are consistent with the numbering of figures and tables in the online appendix. c) "Figures": This is where all figures used in the online appendix are saved in pdf format (empty until the program is executed) d) "Output": Measures of relative usefulness, true positive rates, false positive rates are saved in a .csv file in this folder. It also contains an .xls file that is automatically updated to produce the panels shown in figure F2 e) "OxCodes for individual unit root tests": This subfolder contains all that is needed to replicate the country-specific unit root (ADF) tests. The folder consists of three subfolders: -- "Codes": Constains Ox codes to conduct country specific unit root tests. Edit line 22 to change the directory (the "root") in "main.ox" to execute. The different options (selection of number of countries, number of variables, max lag for AIC etc.) are explained in the code. "main.ox" calls the other codes that performs and stores output from country-specific unit root tests, where lag length is selected by minimizing AIC. -- "Data": The data series to be tested for each country are saved in "Country_data.xlsx", which is called by the Ox program. The numbering of the different sheets is consistent with the "country_id" string provided in the .dta and .csv files described above. "Data.csv" is just a help file to tell the program what time period we are considering, and that the data are at the quarterly frequency. -- "Output": The file "staionarity_fractions.xls" constains the fraction of times where the I(0) null cannot be rejected, while "Table B1.xlsx" automatically organizes the results into the same structure as those reported in Table B1 (note, also the results of the panel unit root tests are reported here) f) "Tables": LaTex tables reported in the online appendix that summarize the results reported in the online appendix (empty until the program is executed) All codes and data will also be made available on http://www.andre-anundsen.com/. Please address any questions regarding replication files and data to: André K. Anundsen Norges Bank Research, Norges Bank, Bankplassen 2, P.O. Box 1179 Sentrum, NO-0107 Oslo, Norway. Or by email to: andre-kallak.anundsen [AT] norges-bank.no.