Strathmore

Defintion of Key words

Total Debt Service (TDS) is the sum of amortization (AMT) and interest payments (INT).
PPG (Public and Publicly Guaranteed):This denotes debt that is either:

Public Debt: Debt borrowed directly by the government or government agencies.

Publicly Guaranteed Debt (PGD): Debt for which a public entity (such as the government) has guaranteed repayment.
IMF repurchases and charges (TDS, current US$): Total repayments made by countries to the IMF, including repurchases and charges.

Abstract

This study investigates the relationship between various debt-related economic factors and suicide rates across different countries, regions, and years. Using a linear mixed-effects model, we analyze the impact of DsePPGTDS (Debt service on external debt public and publicly guaranteed PPG (TDS)), DsExt (Debt service on external debt total (TDS)), IMFrepTDS (IMF repurchases and charges (TDS)), and MultTDS (Multilateral debt service (TDS)) on suicide rates. The results indicate that DsePPGTDS and DsExt significantly influence suicide rates, with higher values associated with lower suicide rates.

This suggests that economic factors related to debt management may positively impact mental health outcomes. Conversely, IMFrepTDS and MultTDS do not show significant effects on suicide rates. The analysis highlights the importance of considering random effects, showing significant variability in suicide rates across different contexts.

The study concludes with recommendations for policymakers to focus on economic support, targeted interventions, further research, and a holistic approach to address mental health issues effectively.

Introduction

Background of the study

This report aims to investigate the influence of Multilateral and bilateral debt and its influence on suicide rates in Southern and Eastern Africa regions.

Focus will be laid on the strength and relationship of multilateral and bilateral borrowing from institutions like the IMF,World Bank or at Bilateral level on suicide rate (per 100 000 incarcerated persons per year).

Studies have shown that Total public and publicly guaranteed debt has been predominantly official (multilateral and bilateral) and concessional, in 1995 Kenya for instance Official debt made up 81 percent of total outstanding debt, and 76 percent of this debt was on concessional terms( loans with favorable conditions such as lower interest rates and longer repayment periods compared to market terms). (Muchabaiwa 2021)

In 2001,Countries in Southern and Eastern Africa experienced relatively high debt distress levels,with the composition of debt shifting from largely concessional loans provided by bilateral and multilateral development agencies to private loans, which made borrowing more expensive.(Ng’eno 2000)

The pressure to repay debt has already resulted into some governments in Eastern and Southern Africa (ESA )to introduce fiscal austerity measures aimed at containing spending and increasing revenue,there has been rapid response initiatives by religious councils seeking reform in the global financial system particulary to G20,G7 and United nations particularly seeking debt forgiveness for african countries.(Nyaga 2024)

Such austerity measures manifested through Unpopular finance bills such as that of Kenya 2024 has sparked social unrest in the region to neighboring countries such as Malawi, Ghana, and Uganda and Tanzania through the GenZ movement decrying for poor public finance management,corruption and bad governance.

Over the years countries in southern and east africa region have been experiencing relatively high suicide rates to the tune of 20 per 100000 people,Suicide is a significant global health concern, with over 703,000 people dying annually (Global suicide statistics).

Economic and social factors are known to influence suicide rates, yet there is limited research on the specific impact (direction and strength) of Bilateral and multilateral fiscal debt metrics on self inflicted harm.

This research therefore aims to explore the relationship between IMF fiscal metrics such as repurchases and charges, multilateral debt service, and various debt service metrics (e.g., total external debt and public and publicly guaranteed debt) with suicide rates.

The following notations will be used to represent the fiscal debt metrics :

DsePPGTDS = Debt service on external debt public and publicly guaranteed PPG (TDS)
DsExt = Debt service on external debt total (TDS)
IMFrepTDS = IMF repurchases and charges (TDS)
MultTDS = Multilateral debt service (TDS)

This research therefore aims to:

Establish the relationship between IMF repurchases (IMFrepTDS) and suicide rates.
Examine the relationship between multilateral debt service (MultTDS) on suicide rates.
Examine the relationship between Total external debt and public and publicly guaranteed debt(DsePPGTDS) on suicide rates and finally
Examine the relationship between suicide rates and Debt service on external debt total (TDS) (DsExt).

Hypotheses

Establish the relationship between IMF repurchases (IMFrepTDS) and suicide rates.

Hypothesis 1:

Null Hypothesis ($H_0$): There is no significant relationship between IMF repurchases (IMFrepTDS) and suicide rates.
Alternative Hypothesis ($H_a$): There is a significant relationship between IMF repurchases (IMFrepTDS) and suicide rates.

Examine the relationship between multilateral debt service (MultTDS) and suicide rates.

Hypothesis 2:

Null Hypothesis ($H_0$): There is no significant relationship between multilateral debt service (MultTDS) and suicide rates.
Alternative Hypothesis ($H_a$): There is a significant relationship between multilateral debt service (MultTDS) and suicide rates.

Examine the relationship between Total external debt and public and publicly guaranteed debt (DsePPGTDS) and suicide rates.

Hypothesis 3:

Null Hypothesis ($H_0$): There is no significant relationship between Total external debt and public and publicly guaranteed debt (DsePPGTDS) and suicide rates.
Alternative Hypothesis ($H_a$): There is a significant relationship between Total external debt and public and publicly guaranteed debt (DsePPGTDS) and suicide rates.

Examine the relationship between suicide rates and Debt service on external debt total (TDS) (DsExt).

Hypothesis 4:

Null Hypothesis ($H_0$): There is no significant relationship between Debt service on external debt total (TDS) (DsExt) and suicide rates.
Alternative Hypothesis ($H_a$): There is a significant relationship between Debt service on external debt total (TDS) (DsExt) and suicide rates.

Methods

Exploratory data analysis methods, including bivariate boxplots analysis, will be employed to understand the distribution of the debt and suicide rates across countries based on the aforementioned debt metrics.

Correlation matrices and plots will be used to understand the nature of the bivariate relationships as well as multicollinearity.

Since the data is longitudinal in nature there will be need to check for correlation between debt metrics variables this inspection will help us in model specification.

Subsequently , linear mixed effects models will be used to estimate the strenght of the relationship between suicide rates (dependent variable) and debt metrics (independent variables). Random effects will include region,country and year, while fixed effects will focus on the four debt service metrics.

Analysis

library(ggplot2)
library(MVA)

## Loading required package: HSAUR2

## Loading required package: tools

library(tidyr)

## 
## Attaching package: 'tidyr'

## The following object is masked from 'package:HSAUR2':
## 
##     household

library(nlme)
library(mclust)

## Package 'mclust' version 6.1.1
## Type 'citation("mclust")' for citing this R package in publications.

library(dplyr)

## 
## Attaching package: 'dplyr'

## The following object is masked from 'package:nlme':
## 
##     collapse

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

library(factoextra)

## Welcome! Want to learn more? See two factoextra-related books at https://goo.gl/ve3WBa

library(knitr)
library(readr)
library(lme4)   # For linear mixed effects model

## Loading required package: Matrix

## 
## Attaching package: 'Matrix'

## The following objects are masked from 'package:tidyr':
## 
##     expand, pack, unpack

## 
## Attaching package: 'lme4'

## The following object is masked from 'package:nlme':
## 
##     lmList

library(lmerTest) # For p-values

## 
## Attaching package: 'lmerTest'

## The following object is masked from 'package:lme4':
## 
##     lmer

## The following object is masked from 'package:stats':
## 
##     step

library(gridExtra)

## 
## Attaching package: 'gridExtra'

## The following object is masked from 'package:dplyr':
## 
##     combine

install.packages("corrplot")

## Installing package into '/cloud/lib/x86_64-pc-linux-gnu-library/4.4'
## (as 'lib' is unspecified)

library(corrplot)

## corrplot 0.92 loaded

# Read your data
data <- read.csv("/cloud/project/8411413/MASTER_DATA_project.csv")
data<-na.omit(data)

# Convert columns to numeric
data$DebtserviceonexternaldebtpublicandpubliclyguaranteedPPGTDS <- as.numeric(data$DebtserviceonexternaldebtpublicandpubliclyguaranteedPPGTDS)

## Warning: NAs introduced by coercion

data$DebtserviceonexternaldebttotalTDS <- as.numeric(data$DebtserviceonexternaldebttotalTDS)

## Warning: NAs introduced by coercion

data$IMFrepurchasesandchargesTDS <- as.numeric(data$IMFrepurchasesandchargesTDS)

## Warning: NAs introduced by coercion

data$MultilateraldebtserviceTDS<-as.numeric(data$MultilateraldebtserviceTDS)

## Warning: NAs introduced by coercion

#Renaming columns
data <- data %>%
  rename(
    DsePPGTDS = DebtserviceonexternaldebtpublicandpubliclyguaranteedPPGTDS,
    DsExt = DebtserviceonexternaldebttotalTDS,
    IMFrepTDS = IMFrepurchasesandchargesTDS,
    MultTDS = MultilateraldebtserviceTDS
  )

# Clean data for each region
clean_data <- function(data, region) {
  data %>%
    filter(region == region) %>%
    select(-SpatialDim) %>%
    drop_na()
}

# Define regions
regions <- list(
  Southern_Africa = c("ZAF", "BWA", "LSO", "SWZ", "NAM", "ZWE", "MOZ", "AGO", "MWI"),
  Northern_Africa = c("DZA", "EGY", "LBY", "MAR", "SDN", "TUN"),
  Eastern_Africa = c("BDI", "COM", "DJI", "ERI", "KEN", "MDG", "MWI", "MUS", "REU", "RWA", "SOM", "SSD", "TZA", "UGA", "ZMB", "ZWE"),
  Central_Africa = c("AGO", "BDI", "CAF", "CMR", "CGO", "COD", "GNQ", "GAB", "SAO"),
  Western_Africa = c("BFA", "BEN", "CIV", "CPV", "GMB", "GHA", "GNB", "LBR", "MLI", "NGA", "SEN", "SLE", "TGO")
)

# Add a region column to the data
data <- data %>%
  mutate(region = case_when(
    Country %in% regions$Southern_Africa ~ 'Southern Africa',
    Country %in% regions$Northern_Africa ~ 'Northern Africa',
    Country %in% regions$Eastern_Africa ~ 'Eastern Africa',
    Country %in% regions$Central_Africa ~ 'Central Africa',
    Country %in% regions$Western_Africa ~ 'Western Africa',
    TRUE ~ 'Other'
  ))

# cleaned_na_data <- clean_data(data, 'Northern Africa')
# cleaned_ea_data <- clean_data(data, 'Eastern Africa')
# cleaned_ca_data <- clean_data(data, 'Central Africa')
# cleaned_wa_data <- clean_data(data, 'Western Africa')
cleaned_sa_data <- clean_data(data, 'Southern Africa')

Conducting summary statistics/Profile Analysis

# Calculate summary statistics
summary_stats <- data %>%
  select(Suicide_Value, DsePPGTDS, DsExt, IMFrepTDS, MultTDS, region, Year) %>%
  group_by(region, Year) %>%
  summarise(
    across(where(is.numeric), list(mean = ~ mean(.)), .names = paste("{col}"))
  ) %>% drop_na()%>%
  rename_with(~ gsub("_mean", "", .))

## `summarise()` has grouped output by 'region'. You can override using the
## `.groups` argument.

# Create kable table
summary_stats_sa<-summary_stats%>%filter(region%in% ('Southern Africa') | region%in% ('Eastern Africa'))
kable(summary_stats_sa)

region	Year	Suicide_Value	DsePPGTDS	DsExt	IMFrepTDS	MultTDS
Eastern Africa	2000	10.184775	115005806	140038171	10344282	36991312
Eastern Africa	2001	9.484297	74276206	97521454	9223498	29653262
Eastern Africa	2002	9.241892	85203448	108693348	10832400	28115325
Eastern Africa	2003	8.570586	115902900	187032232	12135832	33182925
Eastern Africa	2004	8.450508	80433517	159795089	30920405	40945534
Eastern Africa	2005	8.389200	98132723	176643614	16897391	58991246
Eastern Africa	2006	8.301772	83505178	192672196	2726851	47256868
Eastern Africa	2007	8.064297	67264309	205359772	2437463	40401866
Eastern Africa	2008	7.629386	67244277	185189043	1908933	30603853
Eastern Africa	2009	7.658542	64163506	246473968	3438713	31186646
Eastern Africa	2010	7.486869	65283969	267775500	3331448	30671786
Eastern Africa	2014	7.207472	174662816	742219783	13672634	45137229
Eastern Africa	2015	6.900903	153648540	566507962	19065172	44504874
Eastern Africa	2016	6.898589	203379758	482182157	19962399	47879916
Eastern Africa	2017	7.019886	266192945	577798818	25488600	74592276
Eastern Africa	2018	6.925278	404541838	792765579	27639846	103980470
Eastern Africa	2019	27.580920	580166497	1050935571	25836441	129550186
Southern Africa	2000	20.182604	559237214	813332186	16594133	40423443
Southern Africa	2001	20.272333	643187959	914470138	4940304	21233982
Southern Africa	2002	21.008179	444364296	902652088	4619167	19350170
Southern Africa	2003	21.611858	508763777	812250382	4933605	20751339
Southern Africa	2004	22.116917	521349830	834321532	7907952	23320331
Southern Africa	2005	22.432738	866122643	1057019904	28049011	27506258
Southern Africa	2006	23.502488	1113786514	1629198548	10503033	25793792
Southern Africa	2007	24.981987	923285922	1324552891	4626344	21192306
Southern Africa	2008	27.372046	621495343	1187991872	1923404	22848674
Southern Africa	2009	27.905350	868301813	1306756442	2091129	25273975
Southern Africa	2010	27.363725	725709070	1298529845	3801150	30085364
Southern Africa	2011	26.737071	856904536	1560789474	3610938	32397694
Southern Africa	2012	27.971475	1322263046	2243739852	4247306	37381036
Southern Africa	2014	27.868613	1514269556	2493372937	68047023	49622666
Southern Africa	2015	27.204492	2907169319	4727541302	51357843	97549878
Southern Africa	2016	25.819050	1975066822	3502621965	34530965	108603273
Southern Africa	2017	24.180267	1809225867	3315263994	11522063	126650496
Southern Africa	2018	23.427688	3229279808	5496543758	9576966	140242885
Southern Africa	2019	54.235041	2345170668	4512364394	16491772	145543244

# Reshape data to long format using reshape2
long_summary_stats <- reshape(
  summary_stats,
  varying = list(
    c("DsePPGTDS", "DsExt", "IMFrepTDS", "MultTDS", "Suicide_Value")
  ),
  v.names = "Value",
  timevar = "Variable",
  times = c("DsePPGTDS", "DsExt", "IMFrepTDS", "MultTDS", "Suicide_Value"),
  idvar = c("Year", "region"),
  direction = "long"
)

summary_stats_sa<-summary_stats%>%filter(region%in% ('Southern Africa') | region%in% ('Eastern Africa'))
# Reshape data to long format using reshape2

long_summary_stats <- reshape(
  summary_stats_sa,
  varying = list(
    c("DsePPGTDS", "DsExt", "IMFrepTDS", "MultTDS", "Suicide_Value")
  ),
  v.names = "Value",
  timevar = "Variable",
  times = c("DsePPGTDS", "DsExt", "IMFrepTDS", "MultTDS", "Suicide_Value"),
  idvar = c("Year", "region"),
  direction = "long"
)

ggplot() +
  geom_line(data = subset(long_summary_stats, Variable != "Suicide_Value"), aes(x = Year, y = Value, color = Variable), linewidth = 1) +
  facet_wrap(~ region, ncol = 2) +  # Arrange the plots in two columns for better spacing
  labs(
    title = "Average Debt summary statistics over time",
    x = "Year",
    y = "Value($)",
    color = "Variable"
  ) +
  theme_minimal() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.title.y.right = element_text(color = "blue", size = 12),
    axis.text.y.right = element_text(color = "blue", size = 10),
    axis.title.y = element_text(color = "black", size = 12),
    axis.text.y = element_text(color = "black", size = 10),
    strip.text = element_text(size = 12),
    axis.title = element_text(size = 12),
    axis.text = element_text(size = 10),
    plot.title = element_text(size = 16, face = "bold"),
    legend.position = "bottom",  # Move legend to bottom to avoid overlapping with plots
    legend.title = element_text(size = 12),
    legend.text = element_text(size = 10),
    legend.key = element_rect(fill = "white", color = NA),
    legend.background = element_rect(fill = "white", color = "grey80"),
    plot.margin = margin(10, 10, 10, 10)  # Add margin around the plot
  ) +
  scale_color_manual(values = c("DsePPGTDS" = "blue", "DsExt" = "green", "IMFrepTDS" = "red", "MultTDS" = "purple"))

ggplot() +
  geom_line(data = subset(long_summary_stats, Variable == "Suicide_Value"), aes(x = Year, y = Value, color = Variable), size = 1, linetype = "solid") +
  facet_wrap(~ region, ncol = 2) +  # Arrange the plots in two columns for better spacing
  labs(
    title = "Trends in Suicide Values by Region Over Time",
    x = "Year",
    y = "Suicide Value",
    color = "Legend"
  ) +
  theme_minimal() +
  theme(
    panel.grid.major = element_blank(),  # Remove major grid lines
    panel.grid.minor = element_blank(),  # Remove minor grid lines
    axis.title.x = element_text(color = "black", size = 12),
    axis.text.x = element_text(color = "black", size = 10),
    axis.title.y = element_text(color = "black", size = 12),
    axis.text.y = element_text(color = "black", size = 10),
    strip.text = element_text(size = 12, face = "bold"),
    plot.title = element_text(size = 16, face = "bold", hjust = 0.5),
    legend.position = "bottom",  # Move legend to bottom to avoid overlapping with plots
    legend.title = element_text(size = 12),
    legend.text = element_text(size = 10),
    plot.margin = margin(15, 15, 15, 15)  # Add margin around the plot
  ) +
  scale_color_manual(values = c("Suicide_Value" = "red")) +  # Customize color for better visibility
  scale_y_continuous(limits = c(0, 59)) +  # Set y-axis limits between 0 and 20
  theme(
    panel.spacing = unit(1, "lines"),  # Add spacing between facets
    legend.key = element_rect(fill = "white", color = NA),  # Improve legend key appearance
    legend.background = element_rect(fill = "white", color = "grey80")  # Add background and border to legend
  )

## Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
## ℹ Please use `linewidth` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.

# Reshape data to long format
long_summary_stats <- summary_stats %>%
  pivot_longer(
    cols = c(DsePPGTDS, DsExt, IMFrepTDS, MultTDS, Suicide_Value),
    names_to = "Variable",
    values_to = "Value"
  )

Discussion

Eastern Africa

Trends in Suicide Value:

The mean Suicide_Value in Eastern Africa showed a decreasing trend from the year 2000 to 2019. It started at 10.18 (per 100000) in 2000 and decreases to 6.93 in 2018, with a slight increase to 7.58 in 2019 probably due to COVID-19 pandemic.

These results however indicate a general decline in the mean Suicide_Value over time, suggesting an improvement or change in the factors influencing suicide rates in the East african region.

Other Variables:

DsePPGTDS Debt service on external debt public and publicly guaranteed PPG (TDS) Showed a varied trend with some fluctuations but does not demonstrate a clear upward or downward trajectory.
DsExt (Debt service on external debt total (TDS), IMFrepTDS (IMF repurchases and charges (TDS)), and MultTDS(Multilateral debt service (TDS)) also exhibited variability but without a strong trend over the years.

This variability and unclear tend likely represent volatile nature of East african countries economic or social indicators that may impact or correlate with Suicide_Value.

Southern Africa

Trends in Suicide Value: - The mean Suicide_Value in Southern Africa started at 20.18 (per 100000) in 2000 and took an overall upward trend, reaching as high as 27.97 in 2012 before slightly decreasing to 24.18 in 2017. This indicates a general increase in the mean Suicide_Value over time, which might suggest worsening conditions or increasing reporting of suicides in the region.

Other Variables: - DsePPGTDS, DsExt, IMFrepTDS, and MultTDS also showed varying trends over the years. For instance, DsePPGTDS and DsExt exhibit significant fluctuations, possibly reflecting changes in economic conditions or data reporting practices. The variability in these indicators might be influencing the Suicide_Value differently across the years.

Summary

Eastern Africa: The Suicide_Value has generally decreased over time, which may indicate improvements in social conditions, healthcare, or reporting practices. Other variables show variability without strong trends.
Southern Africa: There is a noticeable increase in the Suicide_Value over time, suggesting potential worsening of conditions related to mental health or changes in reporting practices. The variability in other variables may be contributing to this trend.

Overall, the data reflects complex interactions between suicide rates and various socio-economic factors. Continuous monitoring and further investigation are necessary to understand the underlying causes and to develop effective interventions.

# Select the relevant variables and clean the data
clean_data <- data %>% 
  select(Suicide_Value, DsePPGTDS, DsExt, IMFrepTDS, MultTDS) %>% 
  mutate(across(everything(), as.numeric)) %>% 
  drop_na()

# Compute the correlation matrix
correlation_matrix <- cor(clean_data, use = "complete.obs")

# Display the correlation matrix
print(correlation_matrix)

##               Suicide_Value  DsePPGTDS      DsExt  IMFrepTDS     MultTDS
## Suicide_Value    1.00000000 0.04123163 0.04124235 0.01600256 -0.06639938
## DsePPGTDS        0.04123163 1.00000000 0.82091981 0.20011419  0.67174749
## DsExt            0.04124235 0.82091981 1.00000000 0.20969330  0.64790910
## IMFrepTDS        0.01600256 0.20011419 0.20969330 1.00000000  0.30329504
## MultTDS         -0.06639938 0.67174749 0.64790910 0.30329504  1.00000000

# Plot the correlation matrix using corrplot with enhanced aesthetics
corrplot(correlation_matrix, 
         method = "color",               # Color method for visualization
         col = colorRampPalette(c("blue", "white", "red"))(200),  # Custom color palette
         type = "full",                  # Display full matrix
         addCoef.col = "black",          # Color for correlation coefficients
         tl.col = "black",               # Color of text labels
         tl.srt = 45,                    # Rotate text labels for better readability
         number.cex = 0.7,               # Size of correlation coefficients
         diag = FALSE,                   # Hide diagonal
         title = "Correlation Matrix",   # Title of the plot
         mar = c(0, 0, 1, 0))            # Margin adjustments

# Function to create bivariate boxplots
# Install and load required packages
install.packages("MVA")

## Installing package into '/cloud/lib/x86_64-pc-linux-gnu-library/4.4'
## (as 'lib' is unspecified)

install.packages("gridExtra")

## Installing package into '/cloud/lib/x86_64-pc-linux-gnu-library/4.4'
## (as 'lib' is unspecified)

library(MVA)
library(gridExtra)

clean_data <- data %>% 
  select(Suicide_Value, DsePPGTDS, DsExt, IMFrepTDS, MultTDS) %>% 
  mutate(across(everything(), as.numeric)) %>% 
  drop_na()

# Define the variable pairs
variable_pairs <- list(
  c("DsePPGTDS", "Suicide_Value"),
  c("DsExt", "Suicide_Value"),
  c("IMFrepTDS", "Suicide_Value"),
  c("MultTDS", "Suicide_Value")
)

# Function to create bivariate boxplot for a pair of variables
create_bivariate_boxplot <- function(data, col1, col2) {
  subset_data <- data %>% select(all_of(c(col1, col2)))
  bvbox(subset_data, mtitle = paste("Bivariate Box plot of", col2, "and", col1), 
        xlab = col1, ylab = col2, col = 'red')
}

par(mfrow = c(2, 2))


## Generate and display the bivariate boxplots
for (pair in variable_pairs) {
  create_bivariate_boxplot(clean_data, pair[1], pair[2])

}

Brief Discussion of the Correlation Matrix

The correlation matrix provides insights into the relationships between the variables in the dataset:

Suicide_Value and Other Variables:
- DsePPGTDS Debt service on external debt public and publicly guaranteed PPG (TDS) (0.041): There is a very weak positive correlation between Suicide_Value and DsePPGTDS, indicating that changes in DsePPGTDS have a positive effect on Suicide_Value.
- DsExt Debt service on external debt total (TDS) (0.041): Similar to DsePPGTDS, there is a weak positive correlation between Suicide_Value and DsExt.
- IMFrepTDS IMF repurchases and charges (TDS) (0.016): The correlation between Suicide_Value and IMFrepTDS is also weak , suggesting that IMFrepTDS has little to no effect on Suicide_Value.
- MultTDSMultilateral debt service (TDS) (-0.066): There is a very weak negative correlation between Suicide_Value and MultTDS, indicating a slight inverse relationship.
High Correlations Among Predictors:
- DsePPGTDS and DsExt (0.821): There is a strong positive correlation between these two variables, which means they tend to increase together. This high correlation may indicate redundancy or multicollinearity and hence need to adopt LME in modelling the effect of Fiscal debt on suicide.
- DsePPGTDS and MultTDS (0.672): Similarly, DsePPGTDS and MultTDS have a strong positive correlation.
- DsExt and MultTDS (0.648): DsExt and MultTDS also show a strong positive correlation.
Other Relationships:
- IMFrepTDS: This variable shows moderate correlations with DsePPGTDS (0.200), DsExt (0.210), and MultTDS (0.303). It indicates some degree of association but not as strong as with other variables.

Handling Multicollinearity with LME Models

Linear Mixed Effects (LME) Models:

Multicollinearity was observed between independent variables/debt metrics, This can lead to issues with estimating the coefficients reliably and interpreting their significance. To ensure efficient estimation of the estimates the LME model was used ### Linear Mixed-Effects Model Notation

The linear mixed-effects model can be described as follows:

Model Notation:

\[ \text{Suicide\_Value}_{ijk} = \beta_0 + \beta_1 \text{DsePPGTDS}_{ijk} + \beta_2 \text{DsExt}_{ijk} + \beta_3 \text{IMFrepTDS}_{ijk} + \beta_4 \text{MultTDS}_{ijk} + u_i + v_j + w_k + \epsilon_{ijk} \]

Where:

$\text{Suicide\_Value}_{ijk}$ is the suicide rate for country $i$, region $j$, and year $k$.
$\beta_0$ is the intercept or population baseline suicide rate .
$\beta_1, \beta_2, \beta_3, \beta_4$ are the fixed effect coefficients for the predictors DsePPGTDS, DsExt, IMFrepTDS, and MultTDS, respectively.
$u_i$ is the random effect for the $i$-th country (country_factor).
$v_j$ is the random effect for the $j$-th region (region_factor).
$w_k$ is the random effect for the $k$-th year.
$\epsilon_{ijk}$ is the residual error term.

Explanation:

The model aims to predict the suicide rates ($\text{Suicide\_Value}$) based on four predictors: DsePPGTDS, DsExt, IMFrepTDS, and MultTDS.
The model includes random intercepts for country, region, and year to account for the hierarchical structure and potential variability at these levels.
By including these random effects, the model acknowledges that observations within the same country, region, or year may be more similar to each other than to observations from different countries, regions, or years.
- Random Effects: LME models include random effects to account for variability across different levels of grouping factors (e.g., country, region, year).

These random effects can help mitigate some of the issues caused by multicollinearity among fixed effects because the random effects are estimated separately from the fixed effects.

Model Specification: By incorporating random effects, the model can isolate the variability attributable to different groups and better handle multicollinearity among fixed effects.
Partial Pooling: LME models use partial pooling, meaning that the estimates for each group are “shrunk” towards the overall mean. This approach helps to stabilize estimates when predictors are highly correlated.
Variance Components: LME models decompose the variance into components attributable to fixed effects, random effects, and residual variance.

This decomposition can help understand how multicollinearity affects the fixed effects.

In summary, while high correlations among predictor variables (multicollinearity) can be problematic for fixed-effects models, LME models can handle these issues more gracefully by incorporating random effects. This allows the model to account for variability at different levels and provides a more robust framework for analyzing complex data with correlated predictors.

# Fit linear mixed effects model

clean_data<-data%>%mutate(country_factor=as.factor(Country))

clean_data <- clean_data %>%
  drop_na(Suicide_Value, DsePPGTDS, DsExt,IMFrepTDS , 
          MultTDS, country_factor)

clean_data <- clean_data %>%
  mutate(across(c(Suicide_Value, DsePPGTDS, DsExt,IMFrepTDS , 
          MultTDS), as.numeric),
         country_factor = as.factor(Country),
         region_factor = as.factor(region))

model <- lmer(Suicide_Value ~ DsePPGTDS+
                      DsExt +
                      IMFrepTDS +
                      MultTDS + 
                      (1 | country_factor) + 
                      (1 | region_factor)+
                      (1 | Year),
              data = clean_data)

## Warning: Some predictor variables are on very different scales: consider
## rescaling
## Warning: Some predictor variables are on very different scales: consider
## rescaling

# Summarize the model
summary(model)

## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula: Suicide_Value ~ DsePPGTDS + DsExt + IMFrepTDS + MultTDS + (1 |  
##     country_factor) + (1 | region_factor) + (1 | Year)
##    Data: clean_data
## 
## REML criterion at convergence: 14247.4
## 
## Scaled residuals: 
##     Min      1Q  Median      3Q     Max 
## -8.5259 -0.2240 -0.0197  0.1716 12.8223 
## 
## Random effects:
##  Groups         Name        Variance Std.Dev.
##  country_factor (Intercept) 73.478   8.572   
##  Year           (Intercept)  6.074   2.465   
##  region_factor  (Intercept) 44.893   6.700   
##  Residual                   16.207   4.026   
## Number of obs: 2390, groups:  country_factor, 123; Year, 20; region_factor, 6
## 
## Fixed effects:
##               Estimate Std. Error         df t value Pr(>|t|)    
## (Intercept)  1.203e+01  3.038e+00  5.003e+00   3.960 0.010733 *  
## DsePPGTDS   -1.552e-10  4.565e-11  2.330e+03  -3.399 0.000687 ***
## DsExt       -3.314e-11  1.154e-11  2.278e+03  -2.871 0.004127 ** 
## IMFrepTDS   -1.169e-11  1.128e-10  2.265e+03  -0.104 0.917422    
## MultTDS     -3.025e-10  2.349e-10  2.334e+03  -1.288 0.197942    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Correlation of Fixed Effects:
##           (Intr) DPPGTD DsExt  IMFTDS
## DsePPGTDS -0.009                     
## DsExt      0.001 -0.575              
## IMFrepTDS -0.002  0.145 -0.004       
## MultTDS   -0.015 -0.165 -0.020 -0.134
## fit warnings:
## Some predictor variables are on very different scales: consider rescaling

# Obtain p-values using lmerTest package
anova(model)

Conclusion

The linear mixed-effects model analysis reveals that two variables, DsePPGTDS (Debt service on external debt public and publicly guaranteed PPG (TDS)) and DsExt (Debt service on external debt total (TDS)), significantly affect suicide rates. Specifically, higher values of these debt indicators are associated with lower suicide rates.

DsePPGTDS: The estimate $\beta$ is -1.552e-10 with a standard error of 4.565e-11, indicating a significant effect (p < 0.001). The confidence interval for this estimate ranges from -2.447e-10 to -6.568e-11.
DsExt: The estimate $\beta$ is -3.314e-11 with a standard error of 1.154e-11, indicating a significant effect (p < 0.01). The confidence interval for this estimate ranges from -5.576e-11 to -1.052e-11.

These results suggest that economic factors related to debt may influence mental health outcomes in the studied regions, possibly through mechanisms such as increased economic support or infrastructure improvements associated with debt management.

On the other hand, IMFrepTDS (IMF repurchases and charges (TDS)) and MultTDS (Multilateral debt service (TDS)) do not show significant effects on suicide rates.

IMFrepTDS: The estimate $\beta$ is -1.169e-11 with a standard error of 1.128e-10, indicating no significant effect (p = 0.917). The confidence interval ranges from -2.315e-10 to 2.081e-10.
MultTDS: The estimate $\beta$ is -3.025e-10 with a standard error of 2.349e-10, indicating no significant effect (p = 0.198). The confidence interval ranges from -7.628e-10 to 1.578e-10.

Recommendations

The study demonstrates that certain debt-related economic factors, specifically DsePPGTDS (Debt service on external debt public and publicly guaranteed PPG (TDS)) and DsExt (Debt service on external debt total (TDS)), have a significant negative association with suicide rates.

This implies that regions with higher levels of debt service on external debt and total external debt tend to have lower suicide rates.

The lack of significant findings for IMFrepTDS (IMF repurchases and charges (TDS)) and MultTDS (Multilateral debt service (TDS)) suggests that these forms of debt repayment do not significantly impact suicide rates in the same way.

The variability in suicide rates across different countries, regions, and years underscores the necessity of including random effects in the model to account for these differences.

Based on the findings, several recommendations can be made for policymakers and stakeholders:

Focus on Economic Support: Efforts should be made to understand the underlying mechanisms by which debt service and external debt contribute to lower suicide rates. This might include improving economic stability, increasing public health funding, or enhancing social services in heavily indebted regions.
Rescaling Predictors: To improve the accuracy of future models, predictors should be rescaled to ensure that they are on comparable scales. This can help in obtaining more precise estimates and better understanding the relationships between variables.
Targeted Interventions: Given the variability in suicide rates across countries, regions, and years, interventions should be tailored to the specific needs and contexts of different regions. Understanding local economic and social conditions can help in designing more effective mental health and economic policies.
Further Research: More research is needed to explore the causal pathways through which debt service and external debt impact mental health outcomes. Longitudinal studies and qualitative research could provide deeper insights into these relationships.
Holistic Approach: Addressing mental health issues requires a holistic approach that includes economic policies, social support systems, and mental health services. Policymakers should integrate these components to create comprehensive strategies for reducing suicide rates.

Actionable Recommendations

Implement Economic Stabilization Programs: Develop and implement programs that provide economic support to regions with high debt service, ensuring that economic stability translates into improved mental health outcomes.
Localized Mental Health Initiatives: Launch mental health initiatives that are tailored to the specific economic and social contexts of each region, considering the unique challenges and needs of the population.
Cross-Sector Collaboration: Encourage collaboration between economic policymakers, mental health professionals, and social service providers to create integrated strategies that address both economic and mental health issues.
Monitoring and Evaluation: Establish robust monitoring and evaluation frameworks to continuously assess the impact of economic policies on mental health outcomes, allowing for data-driven adjustments and improvements.

References

Data sources

OData.Feed(“https://ghoapi.azureedge.net/api”, null, [Implementation=“2.0”])
The Looming Debt Crisis in Eastern and Southern Africa: What it Means for Social Sector Investments and Children

Appendix

Debt distress

Muchabaiwa, Bob Libert. 2021. “The Looming Debt Crisis in Eastern and Southern Africa: What It Means for Social Sector Investments and Children.”

Ng’eno, N. K. 2000. “The External Debt Problem of Kenya.”

Nyaga, Beth. 2024. “African Interfaith Leaders Urge Debt Forgiveness, Economic Reforms.”

Multivariate Analysis Project

Hassan S. Ngala