Firm attributes and business obstacles: Insights from probit regression of World Bank Enterprise Survey data
Description
Research Overview and Hypotheses This repository contains the replication code and methodological documentation for our study examining the relationship between firm attributes (size, age, ownership structure, growth rate, and managerial experience) and the severity of business obstacles in Zambia. Grounded in Institutional Theory, the Resource-Based View (RBV), and Credit Rationing Theory, the research tests the hypothesis that firm characteristics differentiate how businesses perceive and experience business obstacles. The study posits that resource-constrained firms (e.g., SMEs or younger firms) are more likely to report higher severity in obstacles such as access to finance, corruption, and infrastructure deficits compared to their larger, established counterparts. Data Description and Collection The underlying data is drawn from the 2019 World Bank Enterprise Survey (WBES) for Zambia. The dofile provides further details. First, Firm Attributes are constructed, recoding raw inputs into usable metrics for firm age (years), size (employee count categories), domestic ownership percentages, and historical growth rates. Second, Business Obstacles are derived from Likert-scale responses (ranging from 0="No Obstacle" to 4="Very Severe Obstacle"). The code further transforms these into binary variables to isolate "Major" and "Severe" constraints for probit analysis. Notable Findings and Interpretation Descriptive analysis of the dataset reveals that the most critical hurdles facing Zambian enterprises are access to finance, electricity supply, competition from the informal sector, customs regulations, tax rates, corruption, and political instability. The regression outputs, which can be reproduced using the attached code, demonstrate that the burden of these obstacles is not uniformly distributed. The data shows that firm attributes act as significant predictors for the intensity of these constraints. For interpretation, the Ordinal and Binary Probit models reveal the marginal effects of specific characteristics; for instance, a positive coefficient for the "Domestic_ownership" variable regarding "Access to Finance" indicates that wholly locally owned firms are more likely to perceive finance as a severe barrier than firms with some foreign ownership. Usage and Replication This repository provides the complete Stata syntax required to clean the raw data, generate descriptive statistics, and execute the regression models presented in the published article. Users should interpret the output tables as evidence of the varying influence of firm demographics on business operations. To use this code, users must download the raw dataset (Zambia-2019-full data.dta) from the World Bank Enterprise Surveys portal, as licensing restrictions prevent redistribution of the data here. The code is annotated to guide the user through recoding variables, generating the horizontal bar chart, and producing the tables for both ordinal and binary probit regressions.
Files
Steps to reproduce
Data Source and Sampling We derived our findings from the 2019 World Bank Enterprise Survey (WBES) for Zambia. This dataset is a stratified random sample ensuring representativeness across geographic regions, business sectors (manufacturing, retail, services), and firm sizes. The data contains responses from business managers and owners. Analytical Protocol and Software Our analysis was performed using Stata 17. The .do file shared in this repository encapsulates our entire methodological workflow, allowing for exact reproduction of the study's results. The workflow proceeds in three stages: (1) Data Preparation: We cleaned the raw survey data, handling missing values and constructing derived variables for firm attributes (age, size category, domestic ownership, and historical growth). (2) Variable Transformation: We recoded the Likert-scale obstacle ratings into binary variables (Major/Severe vs. Non-Major) and ordinal categories to suit specific regression models. (3) Statistical Modeling: We executed Ordinal Probit and Binary Probit regressions to estimate the marginal effects of firm attributes on the perception of business obstacles. Reproduction This repository contains the code required to generate the study's tables and figures. To reproduce the analysis, researchers must download the raw dataset (Zambia-2019-full data.dta) directly from the World Bank Enterprise Surveys portal and execute the provided .do file.