Digital Capital Gap in Andalusia: A Logistic Regression Approach

Project Overview

This project investigates the factors associated with the possession of technical-digital capital among residents of Andalusia (Spain) using microdata from the 2025 ICT-Households Survey (TIC-Hogares 2025) conducted by the Spanish National Statistics Institute (INE).

The analysis applies binary logistic regression to examine how sociodemographic, economic, labor-market and technological factors influence the probability of possessing technical-digital capital.

The project was developed as part of the MSc in Social Data Science at the University of Granada.

---

Project Highlights

• Analysis based on official microdata from the 2025 ICT-Households Survey (INE).
• Construction of a composite indicator of technical-digital capital.
• Binary logistic regression modelling.
• Estimation of Odds Ratios and Average Marginal Effects.
• Evaluation of predictive performance using ROC curves and AUC.
• Assessment of gender-based interaction effects.
• Identification of educational, technological and labor-related determinants of digital capital.

---

Research Questions

• Which factors are associated with possessing technical-digital capital in Andalusia?
• How do education, employment status and income influence digital capital?
• Does access to digital technologies increase the likelihood of possessing digital capital?
• Are there gender differences in the determinants of digital capital?
• How well can a logistic regression model predict the presence of technical-digital capital?

---

Data Source

The analysis uses microdata from:

ICT-Households Survey 2025 (TIC-Hogares 2025)
Spanish National Statistics Institute (INE)

The original microdata are not distributed in this repository.

Data can be obtained from:

https://www.ine.es/dyngs/Prensa/TICH2025.htm

---

Methodology

The analysis followed five main stages.

1. Data Preparation

• Selection of Andalusia residents.
• Variable recoding and cleaning.
• Construction of the dependent variable:
  • Technical-Digital Capital (1 = possesses digital capital; 0 = otherwise).

2. Exploratory Data Analysis

• Descriptive statistics.
• Frequency distributions.
• Assessment of missing values and variable quality.

3. Logistic Regression Modelling

A binary logistic regression model was estimated using:

• Sociodemographic variables.
• Educational attainment.
• Employment status.
• Household characteristics.
• Access to digital technologies.
• Internet usage patterns.
• Teleworking indicators.
• Gender interaction effects.

4. Model Evaluation

Model performance was assessed using:

• Wald tests.
• Likelihood Ratio tests.
• Hosmer-Lemeshow goodness-of-fit test.
• ROC curve analysis.
• Area Under the Curve (AUC).

5. Effect Interpretation

Results were interpreted through:

• Logistic coefficients.
• Odds Ratios.
• Average Marginal Effects (AMEs).

---

Main Results

• Educational attainment is one of the strongest predictors of technical-digital capital.
• Students exhibit the highest probability of possessing digital capital.
• Frequent internet use substantially increases digital capital acquisition.
• Teleworking is positively associated with technical-digital capital.
• Access to a computer at home significantly improves digital capital outcomes.
• Age is negatively associated with the probability of possessing digital capital.
• Most gender interaction effects are relatively small or statistically insignificant.
• The logistic regression model achieved strong predictive performance:

AUC = 0.865

indicating very good discrimination between individuals with and without technical-digital capital.

---

Repository Structure

digital-capital-gap-andalusia-logistic-regression/
│
├── data/
│   └── README.md
│
├── results/
│   ├── odds-ratios.png
│   ├── marginal-effects.png
│   ├── roc-curve.png
│   ├── logit-model-coefficients.png
│   └── README.md
│
├── Digital_Capital_Gap_Andalusia_Logit_Model.Rmd
├── Digital_Capital_Gap_Andalusia_Logit_Model_Report.pdf
│
├── packages.R
├── Referencias.bib
├── apa_7th.csl
│
└── README.md

---

Technologies Used

• R
• R Markdown
• Tidyverse
• Psych
• Margins
• Car
• pROC
• Caret
• PSCL
• ResourceSelection
• AOD
• Broom
• Xtable

---

Skills Demonstrated

• Logistic Regression
• Statistical Modelling
• Binary Classification
• Predictive Analytics
• Odds Ratio Interpretation
• Marginal Effects Analysis
• Model Validation
• ROC-AUC Evaluation
• Survey Data Analysis
• Data Cleaning and Preprocessing
• Exploratory Data Analysis (EDA)
• Reproducible Research
• Quantitative Social Research
• Digital Inequality Research
• Computational Social Science

---

Author

Carlos F. Carreras De León

MSc in Social Data Science

University of Granada