This section discusses the variables used to proxy risk factors and determine the probability of a project failing. The World Bank's Private Participation in Infrastructure Database is the primary data source for this analysis. The database does not contain all infrastructure projects with private investment, but it has the widest coverage of private investments in developing countries with project-related information. This study excludes full divestitures and merchant projects, whereby private sponsors build new infrastructure in liberalized markets, but get no government revenue or payment guarantees. Based on these criteria, 6,273 PPP projects are considered, including 2,819 PPP projects in developing Asia.
Each project contains the following information: host country, sector of investment, type and degree of private participation, project modality, duration, status, and financing. The project data are complemented with host country macroeconomic, socioeconomic, and political characteristics. The descriptive statistics are in Appendix A3.1. Although the World Bank's Private Participation in Infrastructure Database tracks infrastructure investments from 1960 to 2016, this study uses projects from 1990 to 2015 to maximize compatibility with nonproject data.
The dependent variable is the survival time or duration of PPP infrastructure projects. All projects are grouped into two statuses: survivors (active projects) and failures (projects declared in distress or cancelled). The duration of failures is measured as the number of years from the financial closure year-that is, the year in which private sponsors entered into an agreement to invest funds or provide services-up to the year the project was cancelled or declared in distress. The duration for failures is completed when a project is declared in distress or cancelled.
The duration of survivors is measured by the number of years from the financial closure year up to the end of the measurement period in 2015. Because the duration of survivors has not been completed, and is unknown up to the end of the sample period, durations of survivors are treated as right-censored observations. Of the 2,819 PPP projects in developing Asia, 95% were active, 4% cancelled, and 1% distressed. Cancelled and distressed projects in Asia have an average duration of 4.5 years.
Independent variables include project-and country-specific variables. Control variables for projects include type of projects, project origination, method of awarding contracts, government contract level (national or provincial, for example), direct and indirect government support, level of private participation, and MDB support. For countries, variables include gross domestic product (GDP) growth per capita, terms of trade, trade and debt ratios, political risk, presence of PPP units, and whether there were natural disasters during the study period. These macroeconomic variables provide a measure of a country's ability to withstand internal and external shocks, which create unexpected situations that can force PPP parties to renege on contract obligations.
To assess the political stability of the countries in this study, the PRS Group's International Country Risk Guide's political risk rating data are used.1 Data are annual averages for the duration of a project based on the two components of the guide's political risk rating: law and order, and corruption. The law component measures the strength and impartiality of a country's legal system; the order component assesses observance of the law. The corruption variable is an assessment of corruption in the political system. For both variables, the lower the score, the higher the risk, and vice versa.