Philippine Poverty Statistics Analysis

PAN DE MIC: PHILIPPINE POVERTY THRESHOLD AND INCIDENCE BEFORE AND DURING PANDEMIC

Hi. We are CS .

And this is our project, Poverty Threshold and Incidence . In March 2020, in response to the COVID-19 pandemic, the Philippine government implemented lockdowns with varying stringency across the country, lasting for more than 17 months, making it one of the longest lockdowns worldwide. These measures limited people's movement and affected many livelihoods. The year 2020 marked the country's deepest recession in post-war history. However, it is worth noting that the inflation rate during that time remained stable. Other factors affecting poverty such as employment, the state of the agriculture sector, and population growth, were also affected by the pandemic. This raises the question: how has the COVID-19 pandemic affected poverty in the Philippines? We aim to delve deeper into this issue and determine the pandemic's impact on each province in the country.

DATA SCIENCE TEAM

Danyael

dddelacruz6@up.edu.ph

Sunny

sssy1@up.edu.ph

Kathleen

kgjocson@up.edu.ph

Ieiaiel

igsanceda@up.edu.ph

REFERENCES
1. Chiu, P. D. M. (2021). Why the Philippines' long lockdowns couldn't contain covid-19. BMJ, n2063.↩
2. Debuque-Gonzales, M. (2021). Navigating the COVID-19 Storm: Impact of the pandemic on the Philippine economy and macro responses of government. In Philippine Institute for Development Studies (DP 2021-39). Retrieved April 18, 2024↩

PROJECT OVERVIEW

Research questions
How has COVID affected poverty in the Philippines?
How did COVID-19 impact poverty across various provinces in the Philippines?

Hypothesis
Null Hypothesis
There is no significant relationship between the COVID-19 pandemic and increased poverty rates in the Philippines.

Alternative Hypothesis
The COVID-19 pandemic has a significant impact on increasing poverty rates in the Philippines due to factors such as the abrupt pause in employment, company downsizing, and the implementation of lockdown protocols, which disrupt income-generating opportunities for many households, leading to intensified economic burden on families.

Proposed Solution
Addressing the poverty rates in the Philippines is a tough task. We would first need to identify the factors that are affecting or have affected our poverty rates, particularly in the context of COVID, which is the primary focus of this project. To answer our research questions, we can analyze poverty trends in the Philippines before and after COVID. Next, we can apply a cross-country analysis/comparison with our data and the poverty rates of other Southeast Asian countries to assess how COVID has affected the Philippines compared to others.

DATA COLLECTION

How did we get our data?
The data was obtained through the Philippine Statistics Authority's (PSA) OpenSTAT platform. From their Full Year Poverty Statistics database, various information was taken like the Annual Per Capita Poverty Threshold among Population, the Poverty Incidence among Families, the Poverty Gap, and much more. These statistics are based on the results of Family Income and Expenditure Surveys from the years 2015, 2018, and 2021 (preliminary).

How big is our dataset?
This dataset comprises statistical measures of poverty spanning three years: 2015, 2018, and 2021. For each of these years, data is provided for six different statistics. Additionally, the dataset is segmented by each region and province in the Philippines. In total, the dataset contains 1883 data points.

Dataset Definitions
Annual Per Capita Poverty Threshold (in Php) - minimum income required for a family/individual to meet the basic food and non-food requirements.
Poverty Incidence (%) - number of individuals with income below the per capita poverty thresholds divided by the total number of individuals.
Subsistence Incidence (%) - number of individuals with income below the per capita food thresholds divided by the total number of individuals.
Poverty Gap (%) - the weighted total income shortfall (expressed in proportion to the poverty threshold) of families/ individuals with income below the poverty threshold divided by the weighted total number of families/ individuals

Dataset

The results showed that time is a significant factor in poverty rates, however, COVID-19 did not necessarily have a significant effect on poverty rates.

Further hypothesis tests could be done on this subject as there might not be enough time points to properly measure the impact of COVID-19 on poverty rates (e.g. There is no data for the years 2019 and 2020, where majority of COVID-19 took place.)

Mann-Whitney U-Test

Test Statistic: 3188.5
p-value: 0.5691758788573739

With a p-value greater than 0.05, we fail to reject the null hypothesis: There is no significant difference in poverty rates between 2018 and 2021.

Kruskal-Wallis

Test Statistic: 15.32939795979
p-value: 0.00046909794194

With a p-value less than 0.05, we reject the null-hypothesis: There is a significant difference in poverty levels across the years.

FRIEDMAN

Test Statistic: 67.7044334975
p-value: 1.986868231641e-15

With a p-value less than 0.05, we reject the null-hypothesis: There is a significant difference in poverty levels across the years.

We first checked if our data was normally distributed and has equal variance.

For the Shapiro-Wilk Test, we found out that the 2015 data was normally distributed, however, the 2018 and 2021 data were not normally distributed.

For the Levene Test, we found out that there was unequal variance across the 2015, 2018, and 2021 data samples.

Due to this, non-parametric statistical tests are preferred. We performed the Mann-Whitney U-test for the 2018 and 2021 dataset, and the Kruskal-Wallis and Friedman tests for the entire dataset.

hypothesis testing

Research question 2: How did COVID-19 impact poverty across various provinces in the Philippines?

Research question 1: How has COVID affected poverty in the Philippines?

data visualizations for EACH research question

exploratory data analysis

Although the results suggest that the pandemic did not significantly affect overall poverty rates, this does not imply that there is no impact at all. In the following section, we will delve deeper into how the pandemic influenced poverty across various provinces in the country by identifying distinct clusters through machine learning.

Machine Learning

Clustering

We decided to use a clustering method to group provinces and regions depending on the level of influence COVID-19 had on their poverty rates

Regions

Provinces

Members of Cluster 0 (low impact) are the least affected by COVID-19 in terms of poverty, followed by members of Cluster 1 (medium impact), and then finally members of Cluster 2 (high impact) are the most affected by COVID-19.

Now that we have done exploratory data analysis to get insights from our dataset, and machine learning to identify the clusters, in the next section, we will be discussing deeper the results of these methods.

Findings

Here's what we found and concluded.

Discussion

Our initial goal with this project was to determine if COVID-19 has any impact on the poverty rates in the Philippines. To do so, we decided to check per region and province. Through our analysis, we found that the results of statistical tests indicated a significant influence. Therefore, this can suggest that the pandemic has disrupted economic activities and led to widespread job losses.

If we examine our Nutshell plot, we can observe a "heat map" illustrating poverty rates for the years 2018 and 2021. This visualization reveals a significant shift: in 2018, the area with the lowest poverty rates was the National Capital Region (NCR), but by 2021, this had changed to Siquijor. However Sulu consistently had the highest poverty rates in both 2018 and 2021. This indicates that while some regions have seen improvements, others have remained consistently disadvantaged.

Other than performing statistical tests, we also trained a Machine Learning model that groups the regions and provinces based on the change in poverty statistics among the population. The impact of COVID-19 on each location is measured by getting the difference between 2021 data (during COVID-19) and 2018 data (before COVID-19). Using this difference, each region and province was grouped into one of 3 clusters representing low, moderate, and high impact of COVID-19 in the poverty of the area. Let us look at the characteristics and implications of each cluster for regions and provinces.

Regions
1. Cluster 0 (Red): These regions were able to control poverty rates.
2. Cluster 1 (Blue): These regions had varying results where some regions improved poverty rates, but some got worse.
3. Cluster 2 (Green): These regions were severely affected and showed increase poverty rates
Province
1. Cluster 0 (Red): These provinces were the least affected by COVID-19 and even showed a reduction in poverty levels.
2. Cluster 1 (Blue): These provinces saw a moderate rise in poverty levels caused by COVID-19.
3. Cluster 2 (Green): These provinces saw a significant rise in poverty levels cause by COVID-19.

In total, there are 4 regions that had low COVID-19 impact on poverty, 5 regions that had moderate COVID-19 impact on poverty, and 8 regions that had high COVID-19 impact on poverty.

For the provinces, there are 2 provinces that had low COVID-19 impact on poverty, 53 provinces that had a moderate COVID-19 impact on poverty, and 32 provinces that had high COVID-19 impact on poverty.

Conclusion

Our hypothesis testing revealed significant differences in poverty rates across the years 2015, 2018, and 2021, as determined by the Kruskal-Wallis and Friedman tests. However, the Mann-Whitney U-test comparing data from 2018 and 2021 suggested that the pandemic did not have a significant effect on poverty rates. Despite the lack of a significant difference between the 2018 and 2021 data, we did not want to undermine the pandemic's impact on the lives of the Filipino people. Hence, to gain deeper insights, we employed a machine learning model to identify distinct clusters of regions and provinces. By doing so, we obtained a clearer understanding of which provinces managed the pandemic effectively and which faced greater challenges.

Upon identifying the provinces that experienced low, medium, and high impacts on their poverty rates, we recommend further study of the provinces in each cluster to uncover commonalities, examine their pandemic response strategies, and identify factors contributing to their outcomes. We also urge national and local government units to intensify efforts to assist provinces in the high-impact cluster (Cluster 2). We suggest reviewing the actions taken during the height of the pandemic to understand why some provinces fared worse than others. This will help ensure that all provinces receive equitable attention and support during future national crises like pandemics.

About Us

Here are the Four Members of CS.

Sunny Sy

4th Year BS Computer Science

Danyael Dela Cruz

3rd Year BS Computer Science

Kathleen Jocson

3rd Year BS Computer Science

Ieiaiel Sanceda

4th Year BS Computer Science