Gross Fixed Capital Formation Determinants
Description
Dataset consists of variables for 6 countries in the period 1995-2021. Dependent variable in the panel models is ln (gross fixed capital formation as % of GDP). Independent variables are listed in Stata log file results_models_Stata 17. Data were imported in Stata v. 17.0 from the excel file Gross Fixed Capital Formation determinants.xlsx. Independent variables (gross fixed capital formation determinants) are: 1. interest_rates - long-term interest rates in respective country/year from OECD website database (at the moment, the site and databases are in the process of migration, so it could be more difficult to access the data than it used to be) https://data.oecd.org/interest/long-term-interest-rates.htm 2. consumption_growth - 2-year lagged growth rate of households final consumption from the World Bank database https://data.worldbank.org/indicator/NE.CON.PRVT.KD.ZG 3. ln_exportsGDP - 2-year lagged ln (exports of goods and services as % of GDP) from the World Bank database https://data.worldbank.org/indicator/NE.EXP.GNFS.ZS 4. corporate_savings - 2-year lagged gross savings from non-financial corporations in PPP 2022 billion EUR from the World Inequality Database https://wid.world/data/ 5. profit_rate - 2-year lagged rate of profit (in %) from the database https://dbasu.shinyapps.io/World-Profitability/ used in article "World Profit Rates, 1960–2019" (Basu et al., 2022) The results of estimation showed that the only variable positively contributing to the level of gross fixed capital formation as percent of GDP is household final consumption growth rate. The highest negative impact comes from the exports of goods and services as percent of GDP. Postestimation tests showed the presence of heteroscedasticity (Modified Wald statistic for groupwise heteroscedasticity, Stata command xttest3), 1st order autocorrelation (Wooldridge test for 1st order autocorrelation, Stata command xtserial) and cross-sectional dependence (Breusch-Pagan LM test for independence, Stata command xttest2). Therefore, the preferred models are pooled OLS with Driscoll-Kraay standard errors, FGLS with options correlated panels and common AR(1), as well as PCSE with option common AR(1) (Prais–Winsten regression).
Files
Steps to reproduce
Data were collected from the sources previously mentioned. All variables were listed in source excel file Gross Fixed Capital Formation determinants.xlsx. This file was imported into Stata v. 17.0, which was used for the estimation process (6 different panel models). Models and Stata commands used are given in the log file results_models_Stata 17.smcl. Postestimation results i.e. Stata commands are given in the log file postestimation_results.smcl.