Exploring the 3P Index: Data-Driven Insights for the MENA Team

NOVEMBER (NEXT STEPS)

• Check Markus’ IMF metadataset and gain info about it! IMF WEO Database: Strengths & Limitations for Competitiveness Analysis*

  Strengths

  Comprehensive coverage: Macroeconomic data for nearly all countries (GDP, inflation, trade, fiscal balances, etc.). Regular updates: Released twice a year (April/October) with the latest projections and historical data. Standardized methodology: Consistent and transparent data collection, ensuring comparability across countries. Credible source: Widely trusted by policymakers, researchers, and financial institutions. Macroeconomic focus: Strong on indicators like fiscal deficits, public debt, and current account balances.

  Limitations

  Not competitiveness-specific: Lacks detailed indicators like innovation, labor market flexibility, or business environment quality. Aggregated data: Mostly national-level aggregates, missing sectoral or regional nuances. No non-economic factors: Does not include governance, institutions, or social indicators (e.g., education, health). No composite index: Unlike WEF’s GCI or World Bank’s Doing Business, it does not provide a competitiveness ranking. Projections vs. actuals: Future estimates may differ from real outcomes.
• Adjust WEF file and add new info (change the data sources all together given the unreliability of data sources from WEF?).
• Construct and INTEGRATE new DATASETS: Excel (first in each corresponding folder, then in big dataset) and then Python for REGRESSIONS. Check below for details.
• Take a look at Tina’s protest datasets. (@ Tina: do we have by now indices measuring revolutions, coups d’état, etc.? No, unfortunately we don’t have. I’ll follow up. We also discussed on different indicators (just successful revolutions and coups d’états, or also protests) and if I remember right, @Sergei, you wanted to look into the GDELT and Mass mobilization data Semuhi had suggested. -> 1. GDELT: https://blog.gdeltproject.org/mapping-global-protest-trends-1979-2019-through-one-billion-news-articles/ A open source database of news articles and search using Google's BigQuery platform. 2. Mass mobilization data: https://dataverse.harvard.edu/dataverse/MMdata The Mass Mobilization (MM) data are an effort to understand citizen movements against governments, what citizens want when they demonstrate against governments, and how governments respond to citizens. The project codes protests against governments - the data cover 162 countries between 1990 and March 2020. For each protest event, the project records protester demands, government responses, protest location, and protester identities.)
• When where the shocks for each country / region?
• Changes for provision: was it one sub-index or a shift between different sub-indexes?
• Try to add the data for 2022 for all three Ps for as many countries as possible.

PRE-MEETING AND MEETING 6/8

Data Verification and Integration

• Review Calculations: Verify Ravi's calculations from section 3-2.
• Construct New Datasets: Develop updated datasets in Excel and subsequently in Python for further analysis.

Communication and Collaboration

• WEF Communication: Check email to the World Economic Forum to gather necessary information or data and evaluate other steps.
• Data Sources Information: Obtain detailed information from Markus regarding data sources and adjust a dedicated file for WEF data.

DATA INTEGRATION FOR FUTURE ANALYSIS AND REGRESSIONS

Fiscal and Economic Data Integration

• Fiscal Capacity Data: Integrate data for fiscal capacity indicators, including:
  • Government spending as a percentage of GDP.
  • Total government revenue (consider sources like IMF).
  • Tax revenue as a percentage of GDP.
• Official Development Assistance (ODA): Integrate data on ODA receipts (utilize OECD sources?), focusing on actual receipts (the actual amounts of aid that have been received and recorded by the recipient countries or organizations) rather than pledges (commitments or promises made by donor countries or organizations to provide a certain amount of aid in the future).
• Consumer Price Index: Incorporate consumer price index data to account for inflation (likely available from IMF).
• Lagged GDP Growth: Integrate lagged GDP growth data, specifically using the average of the three previous years.
• Population Metrics: Integrate data for:
  • Logarithm of population.
  • Urbanization level.
  • Population density.
• Age Dependency Rate: Integrate the age dependency rate, defined as the share of people below 15 and above 65 relative to the total population.

Sector-Specific Data

• Resource Receipts: Consider integrating receipts from mineral and energy sales (potential source: IMF).
• Military Spending: Integrate military spending data (consider using Stockholm International Peace Research Institute data?).

Social and Environmental Data

• Conflict Data: No, as it is already covered under the protection index.
• Protest Datasets: Review Tina’s datasets related to protests.
• Migration and Neighboring Influence: Explore data on:
  • Share of immigrants and the influence of proximity to conflict countries.
  • Cooperation between leaders.
  • Dummy variables for neighboring countries in conflict (Tina will provide further insights).
  • Out-migration patterns.
• Natural Disaster Costs: Integrate data on the cost of natural disasters.

BRAINSTORMING INDEPENDENT VARIABLES (PRE-MEETING AND MEETING 10/7)

Suggestions from Amirah with Markus' and Tina's comments:

@ all how about measuring competing positions in government budget such as military, police etc.? Good idea. Police expenditure might also give a hint about repression of protests, see above. We would then need to factor in whether there is or has been in the recent past an open intra-state conflict or civil war.

A. additional regressors/control variables - they should mainly “not” be correlated with our main regressor GDP growth:

1. Official Development Assistance (ODA): in low and middle -income countries, a large share of service delivery maybe funded externally.

Agree! (Markus) yes (Tina)

2. log(population) and the urbanization level: population scale and density or the scale and geography of the population could affect service delivery (e.g. per capita coverage, logistical costs). level of urbanization measured as= (Urban population/ Total population) x 100 Agree with both! (Markus) yes, very good points (Tina)

3. Institutional quality (Rule of Law, Government Effectiveness, Control of Corruption, Democracy Index/regime type: from the World Governance Indicators): strong institutions help translate GDP into actual service delivery. Countries with same income levels may differ a lot in performance due to governance quality and responsiveness.

Agree with Rule of Law, Government Effectiveness and Control of Corruption! In turn, Democracy Index is already included in our Participation Index! (Markus) Plus we had the rule of law index under provision, though we didn’t get enough data for a lot of countries (Tina)

4. Demographics such as % of population under 15, that is the dependency ratio: young populations require more education and health services, influencing both needs and policy priorities (the assumption would be the higher the rate the more we should observe gov provision.

I would suggest to take total age dependency ratio instead as we did in our first article! (Markus) agree (Tina)

5. Fiscal capacity such as government spending as % of GDP, tax revenue as % of GDP: GDP enables capacity, but actual government spending channels that capacity into services (I have doubts about this one but worth it to look into it, I see also the counter logic not to use it).

Correlates too much with our provision index. Government spending is exactly the output that we measure! (Markus) I don’t think that this is exactly what we did before. Fiscal capacity as suggested by Amirah would be the whole cake, and the different provision subindices we measured would be the slices of the cake. For me, it is the same logic as with ODA figures, see above. So far we used high-income vs low-income as a proxy for that (see our first paper) but does this necessarily reflect the fiscal capacity of the government? (Tina)

6. Human capital indicators such as the literacy rate (or illiteracy for that matter) and average years of schooling: These are both outcome variables but are also “facilitators” of service provision, especially in health and education sectors (the higher the easier is distribution so to speak).

Correlates too much with our provision index! (Markus) I’m not sure this brings a lot of new insights. As Markus wrote, it is correlated with our provision index, and the part that is not correlated to it is very differentiated and difficult to argue (household decisions why not to send their kids to school, due to poverty and need of income, social norms, lack of awareness…) (Tina)

7. Conflict such as the presence of conflict, fragility index, or dummy for fragile states: conflict-affected countries likely to experience service delivery breakdowns regardless of GDP. This variable is included in our protection index but not in the provision index so can be safely used as a regressor for the Provision P.

Correlates too much with our protection index index (Markus) I agree (Tina)

BRAINSTORMING FROM MEETING 28/5

• Markus -> We have identified the following possible independent variables as possible explaining ones:

  • Growth (or change in growth)
    • Problem with growth (Amirah): Conceptual and Statistical Challenges
      • Regression Setup:
        1. Index on Growth: Regressing a level variable (like a government service provision index) on a change variable (like annual GDP growth) can be problematic. It assumes that short-term economic fluctuations directly impact structural outcomes, which may not be the case.
        2. Change in Index on Growth: Regressing changes in the index on GDP growth might face issues due to limited variation in the dependent variable and high volatility in GDP growth, potentially leading to weak or insignificant results.
      • Alternative Approaches:
        1. Lagged GDP Growth: Using lagged GDP growth as a predictor can better capture delayed effects on the index level.
        2. Fixed Effects: Incorporating year fixed effects can account for shocks in specific years (e.g., financial crisis, COVID-19). Even though this approach is quite challenging with such a small sample.
  • Revolutions, coups d’état, seizures of power,
  • Wars, civil wars
  • Terrorism
  • Changes in the share of tax revenues as a share of GDP
  • Protests

  Please let us all think about:

  • Possible additional ones (Check Amirah's list above)
  • Indicators to measure these variables

• Summary of Amirah's Guidelines -> Amirah has provided a comprehensive set of guidelines and considerations for conducting regression analysis on your dataset, particularly focusing on the relationship between government service provision indices and economic growth. Here's a breakdown of her points:

  1. Regression Strategy
  • Panel Data: Utilize the panel nature of your dataset with country fixed effects to control for unobserved time-invariant characteristics (Markus' suggestion: do it with and without -> country or region specifics) and year fixed effects to account for time shocks. (Markus' suggestion: do it with and without -> effects of global shocks at specific times).
  • Clustered Standard Errors: Use clustered standard errors at the country level to address serial correlation.
  • Lagged Variables: Consider using lagged GDP growth or average growth over past years to reflect delayed impacts.
  • Incremental Introduction of Controls: Introduce additional control variables one by one to observe their impact on the regression results and ensure robustness.
  2. Insights and Next Steps
  • Regression Implementation:
    1. Model Specification: Start with a basic model and gradually introduce control variables to assess their impact on the regression coefficients and overall model fit.
    2. Diagnostics: Conduct diagnostic tests to check for multicollinearity, heteroskedasticity, and serial correlation. Address any issues that arise.
  • Interpretation and Reporting:
    1. Focus on Robustness: Emphasize the robustness of your findings and be prepared to interpret results even if they are not statistically significant.
    2. Visualization (while estimating causal effects you always summarize your main results in smart and effective pictures): Continue to leverage the visual results you have produced. Highlight interesting trends and outliers in your analysis, as these can provide valuable insights even in the absence of strong regression results.

Remember:

• Variance: Do the three indices display existing disparities between different countries? -> check interval / range.
• Plausibility: Are the results in line with possible explanations? -> values as in the figures in chapter 4.2.
• Consistency between indices: Are the indices scaled in a consistent way? -> compare the 3 Ps respective values as they cover a similar range, and their means and medians are not too different. check also the correlation between each other.
• Consistency within indices: Are the components of the three indices (the values of their aspects) consistent, that is, reasonably correlated? -> check correlations of the components for each 3P index.
• Consistency with other indices: Are the three indices in line with other indices (i.e. do they fulfil the criterion of “concurrent validity”)? -> check consistency with popular indicators such as per capita income and indices such as the HDI or the Global Peace Index.
• Added value: Do the indices add information to other available indicators and indices? -> check which countries are better for each 3P index.

Done

• Countries that have several observations missing and were previously left out: Belarus (left out because data missing for provision index), Eswatini (left out because data missing for protection and provision index), Gabon (left out because data missing for provision index), Iraq (left out because data missing for provision index), Papua New Guinea (left out because data missing for provision index), South Sudan (left out because country has come into being during the period 2007-19), Sudan (left out because data missing for provision index), Uzbekistan (left out because data missing for provision index).
• Created sub-datasets for Conflict/Post-Conflict countries, GCC and Repressive Countries, MENA Region, and EU Countries.
• Added back the eight countries that were left out due to missing data and created a comparison between the two datasets (with and without these countries).
• Plots and correlation analysis for the three indices (Protection, Provision, Participation) and some of their components.
• Checked 2007's protection index (Double checked the excel sheet if the issue about “median/average” has been settled now everywhere in the excel file).

Description

• This project uses the data from my work at IDOS (German Institute of Development and Sustainability) in the Stabilization and Development in the Middle East and North Africa team. My work has been conducted under the supervision of M. Loewe and T. Zintl, co-authors of the discussion paper "Operationalising social contracts: towards an index of government deliverables" (2024) (Further details are available here https://www.idos-research.de/en/mena/ and here https://www.idos-research.de/discussion-paper/article/operationalising-social-contracts-towards-an-index-of-government-deliverables/ ). The research primarily focuses on the themes of Social Protection and Inclusive Growth.
• In recent months, and previously to the beginning of this project, our team has collected and processed data for 154 countries over a 12-year period (2007–2019). This work has resulted in a dataset designed to evaluate each country’s performance concerning social contracts. The dataset is structured to include indicators for the "3P" indexes (Protection, Provision, Participation), with each indicator standardized to a 0–1 scale and weighted based on its relevance to the respective index.
• The list of aspects included in each of the six elements of social contracts: Protection: collective security (i) against foreign threats and (ii) against acts of civil war; security of individuals/citizens (iii) against physical threats such as alleged or real terrorist and criminal acts and (iv) against political threats by own government; (v) human rights aspects of rule of law (including the law as such, especially the existence and enforcement of human and civil rights); and (vi) security against natural, environmental and other macro risks. Provision of economic and social services: (i) infrastructure (communication, information, transport, utilities), (ii) education, (iii) health services, (iv) social protection, (v) poverty reduction, (vi) employment, (vii) economic aspects of rule of law (transparency, fair competition, reliability of government regulation), (viii) a good business climate and (ix) resources in production (e.g. water, land). Participation by society in political decision-making by (i) free, fair and secret elections, (ii) open public debates and (iii) free mass media and other channels. Citizens’ acceptance of the rule of the government. Citizens’ delivery of (i) taxes and other obligations such as (ii) military or civil service, (iii) respect of public order, (iv) engagement in civil society (e.g. neighborly help, support for school child care) or (v) financial donations to social work. Deliverables exchanged between social groups and citizens: (i) mutual respect and recognition, (ii) dialogue on conflictive issues, (iii) mutual support (of course, there is some overlap in contents with engagement and financial donations, mentioned already in the previous element, but they also have an intra-societal specification).
• In particular, here is a detailed overview of the three indices for government deliverables of social contracts:
  • Protection:
    • External threats (Weight: 20.00%, Index: FFP Fragile States Index X1, Source: The Fund for Peace).
    • Civil wars (Weight: 20.00%, Index: UCDP data on fatalities in civil wars, Source: University of Uppsala).
    • Criminal acts (Weight: 20.00%, Index: Global Competitiveness Index Pillar 1 (Security), Source: World Economic Forum).
    • State terror (Weight: 20.00%, Index: Political Terror Scale, Source: University of North Carolina).
    • Rule of law /human rights (Weight: 20.00%, Index: FFP Fragile States Index P3, Source: The Fund for Peace).
    • Environmental threats (Weight: not yet included).
  • Provision:
    • Water, Land (Weight: not yet included).
    • Infrastructure (1. Weight: 6.25%, Index: Global Competitiveness Index Pillar 2 (Transport and utilities), Source: World Economic Forum; 2. Weight: 6.25%, Index: Telecommunication Infrastructure Index, Source: UN Statistics Division).
    • Education (1. Weight: 6.25%, Index: Gov't expenditure on primary and secondary education (% of GDP), Source: World Bank; 2. Weight: 6.25%, Index: Global Competitiveness Index Pillar 6.4 (Skills of future workforce), Source: World Economic Forum).
    • Health (1. Weight: 6.25%, Index: Gov't health expenditure (% of GDP), Source: World Bank; 2. Weight: 6.25%, Index: Out of pocket expenditure (% of total national health care spending), Source: World Bank).
    • Social protection (1. Weight: 6.25%, Index: Public social protection expenditure excl. health (% of GDP), Source: World Bank; 2. Weight: 6.25%, Index: Share of people above retirement age receiving an old-age pension, Source: International Labour Office).
    • Poverty reduction (1. Weight: 6.25%, Index: Public expenditure on social safety nets (% of GDP), Source: World Bank; 2. Weight: 6.25%, Index: Vulnerable persons covered by social assistance (%), Source: International Labour Office).
    • Employment (1. Weight: 6.25%, Index: Share of wage employment on work age population (%), Source: International Labour Office; 2. Weight: 6.25%, Index: Working poverty head-count rate (%), Source: International Labour Office).
    • Rule of law (economic) (1. Weight: 6.25%, Index: Global Competitiveness Index Pillar 1F (Property rights), Source: World Economic Forum; 2. Weight: 6.25%, Index: Global Competitiveness Index Pillar 1E (Incidence of corruption), Source: World Economic Forum).
    • Markets (1. Weight: 6.25%, Index: Global Competitiveness Index Pillar 7A (Market competition), Source: World Economic Forum; 2. Weight: 6.25%, Index: Global Competitiveness Index Pillar 1E (Public-sector performance), Source: World Economic Forum).
  • Participation:
    • V-Dem Index on electoral democracy (Weight: 50.00%, Source: University of Gothenburg).
    • Voice and Accountability Indicator (Weight: 50.00%, Source: World Bank).
• We also realized that it is difficult to find meaningful indicators for the last three elements, that is, the deliverables provided by and among society in general. Some of their aspects are covered by the questions included in the World Value Survey, but only few of their aspects. Other databases, such as the Afrobarometer, include more aspects, but they cover only a limited number of countries. For this reason, the measurement of social cohesion developed by Leininger et al. (2021) currently contains only data on African countries. Therefore, we decided to disregard these three elements in our first move towards measuring social contracts and instead focus fully on the three Ps that governments can give to society. We measure thus just the efforts of one side of the social contract; we cannot yet assess how much these efforts impact and depend on the deliverables of the other parties (e.g. the readiness of society to pay taxes, do military service and contribute to other public goods). We plan to conduct this second step in another paper.

Objectives

Next paper -> development of countries over time, with an obvious focus on MENA countries (trends).

How to run the project

The Data

All the necessary data is present in the data folder in src/idos_ppp. It is possible to download the dataset used to merge the continent variable with the cleaned dataset from https://ourworldindata.org/grapher/continents-according-to-our-world-in-data .

Programs set-up

To set up this project, you first need to install Miniconda and Git. Once those are installed, you can proceed with creating and activating the environment.

Creating and Activating the Environment

Start by navigating to the project's root directory in your terminal, and then type the following into the console:

$ mamba env create -f environment.yml
$ conda activate idos_ppp

Building the Project

The src folder contains all the source code necessary to run this project. Files that start with the prefix task_ are pytask scripts, which execute when you run the following command in the console, as they build up the whole project:

$ pytask

The tests folder includes test scripts that check the functionality of the functions defined in the source code. In order to run them, type:

$ pytest

It is important to run pytask and then pytest, in this order, such that the tests for the plots work.

If you encounter any issues, refer to the sections "Preparing your system" and "How to get started on a second machine" in this website, which is based on the template used for this project.

Credits

The template for this project is from econ-project-templates.

Contributors

@SergeiMolinari