## wpiea2022176-print-pdf — Private Savings and COVID-19 in Sub-Saharan Africa

## Source details

**Canonical URL:** [wpiea2022176-print-pdf — Private Savings and COVID-19 in Sub-Saharan Africa](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022176-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022176-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022176-print-pdf.pdf.json)

---

### Background and objectives
- Low private saving rates in Sub-Saharan Africa (SSA) constrain consumption smoothing, risk management, and financing of life goals and the SDGs.
- Prior to the pandemic private saving rates in SSA had been catching up but remained below other EMDE country groups; heterogeneity is large (notably low in fragile states and volatile, shock-exposed economies).
- Access to formal financial products rose but close to 46 percent of the population is still not able to save at all and less than a third of savers in 2017 saved through formal financial institutions.
- Paper objectives:
  - Reexamine main determinants of private savings in SSA and compare with other world regions.
  - Investigate the impact of COVID-19 (cases, deaths, and preventive measures) on private savings in SSA (period 2017−21).

### Historical and regional trends
- Private saving rate increased to 17.3 percent in 2019 from 11.5 percent in 1983.
- Heterogeneity:
  - Oil exporters and middle-income countries (MICs) are highest savers in the region.
  - Private saving rates are particularly low in fragile states and low-income countries (LICs).
- COVID-19 period aggregate patterns (2020–2021):
  - For the sample of 36 SSA economies, average private saving rates in 2020 and 2021 remain at the same level as 2019: 17.3 percent of GPDI.
  - Median private saving rates fell from 19.1 in 2019 to 18.3 percent of GPDI in 2020, reaching 18.8 percent of GPDI in 2021.
  - Estimated sample of 31 SSA economies: average decline in private savings combined in 2020 and 2021 = -0.76 percent points of GPDI.
  - By contrast: in AEs the increase in private savings during 2020 and 2021 is on average above 5 percent of GPDI; in other EMDEs (excluding SSA) average increase was 1.9 percent of GPDI.
  - Full sample of LICs: average change in private savings for 2020 and 2021 is also negative.

### Micro evidence from UNDP household surveys (six countries, Q2–Q3 2020)
- Countries: Kenya, Nigeria, Rwanda, South Africa, Uganda, Zambia. Sample sizes: 1,159 (Zambia) to 6,120 (Nigeria).
- Key survey findings:
  - A significant share of households depleted savings during COVID-19; savings depletion is first or second source of financing during COVID-19 in all six countries.
  - Savings are main emergency income source for rural and urban households, and across occupations (salaried, self-employed, farming).
  - Depletion of savings frequently linked to transportation and labor supply issues associated with social distancing and lockdowns.
  - Social distancing was the main form of COVID-19 prevention across occupations.
- Selected precise shares (Share of households surveyed, percent — numbers from Figure 5):
  - Kenya (Panel I): Working 14.7, Savings 24.6, Loans 23.8, No source 36.8.
  - Nigeria (Panel II): Working 17.2, Savings 42.2, Loans 19.2, No source 21.4.
  - Rwanda (Panel III): Working 8.8, Savings 31.0, Loans 37.9, No source 22.3.
  - Uganda (Panel IV): Working 12.6, Savings 29.0, Loans 16.0, No source 42.5.
  - South Africa (Panel V): Working 10.9, Savings 22.0, Loans 18.8, No source 48.3.
  - Zambia (Panel VI): Working 32.7, Savings 24.1, Loans 28.8, No source 14.4.
- Micro results corroborate macro findings of lack of increase and small decline in aggregate private savings in SSA during 2020–2021.

### Econometric specification and data
- Baseline dynamic panel (model 1): s_{i,t} = γ s_{i,t−1} + β B_{i,t} + δ X_{i,t} + α_i + τ_t + u_{i,t}, where s_{i,t} = private savings rate (% of GPDI).
- Baseline regressors: lagged dependent variable; ln(real per capita GPDI in PPP); real growth rate of per capita GPDI in PPP; annual inflation; flow of private sector credit (% GPDI); public saving (% GPDI); terms of trade; share of urban population.
- Estimator: primary — Two-stage System GMM (Blundell and Bond, 1998; Roodman, 2006); OLS FE reported for illustration.
- Final panel sizes reported:
  - SSA baseline: at most 31 countries, 986 observations.
  - Full (world) sample: 128 countries, 3,619 observations.
- COVID-19 variables: WHO-based cases and deaths from ourworldindata.org aggregated to annual 2020 and 2021 and expressed per million. Averages (2020–2021) in 13 SSA economies used for some equations: 4,190 cases per million; 59 deaths per million.
- Vaccination: aggregated total vaccine shots 2020–2021 divided by population; average vaccine shots in SSA sample by end-2021 ≈ 5 percent of population.

### Key empirical findings — historical determinants
- Real per capita economic growth:
  - Main robust historical determinant for SSA is real growth rate of per capita GPDI (PPP).
  - 2-Stage System GMM (SSA sample): a one percentage point increase in annual real per capita GPDI growth raises the private saving rate, on average, by 0.46 percentage points (Column (2), Table 1).
  - The paper also reports “Every 1 percentage point increase in real per capita GPDI growth is, on average, associated with an increase in 0.45 percentage points of GPDI in the SSA countries’ private savings rate” in the main text.
  - An increase of 2 percentage points (around the sample average of real per capita GPDI growth between 2020 and 2021 in SSA) would be expected to increase the private savings rate by 0.9 percentage points of GPDI (ceteris paribus).
- Other baseline variables:
  - In baseline specifications lagged dependent variable, rate of urbanization, and public savings are not always significant; in some augmented estimations they are significant with expected signs.
  - Full sample (world) GMM: statistically significant variables include lagged dependent variable, ln(real per capita GPDI), terms of trade, inflation, and public saving/GPDI. Public saving/GPDI coefficient (full sample) = -0.34.

### COVID-19 impacts on private savings — main results
- Specification strategy:
  - First-difference regressions (Equations (3) and (4)) for 2017−2021: Δ(private savings rate) on Δ lagged dependent variable, controls, and annual COVID-19 cases per million or deaths per million.
  - COVID-19 variables treated as strictly exogenous; misreporting likely biases estimates downward (estimates provide lower-bound effects).
- Cases per million:
  - Not statistically significant in preferred 2-Stage System GMM estimations for SSA (Equation (3), Table 3).
  - Some OLS FE specifications show negative associations but GMM does not find significance for SSA.
- Deaths per million:
  - Statistically significant and negative association with change in private savings rate in SSA in 2-Stage System GMM estimations (Equation (4), Table 4).
  - Economic magnitudes:
    - Every 10 cumulative COVID-19 deaths per million inhabitants is associated with a decline of around 0.2 percentage points of GPDI in the change of the private savings rate (ceteris paribus, on average).
    - Example: 30 deaths per million → predicted decline in private savings rate ≈ 0.6 percentage points of GPDI.
    - Using 59 deaths per million (SSA average 2020–2021) implies a decline close to 1.2 percentage points of GPDI in the change of SSA’s private savings rate (ceteris paribus).
- Interpretation: COVID-19 mortality influenced agents toward dissaving in SSA; cases alone not statistically linked to decline in private savings in preferred estimations.

### Stringency and vaccination tests
- Stringency index:
  - Annual stringency measure (Oxford/Hale) not statistically significant for SSA or other samples in 2-Stage System GMM (Table 5). Annual index may not capture forced-savings/foregone-consumption effects.
- Vaccination (total vaccine shots as percent of population, 2020–2021):
  - Not statistically significant in 2-Stage System GMM for SSA or other samples (Table 6).
  - Low rollout: SSA average vaccine shots ≈ 5 percent of population by end-2021 — limited detectability of vaccination effects.
- Combined specifications:
  - In some subsamples (LICs; EMDEs excluding SSA) combining cases with vaccination changes significance and signs of coefficients; for SSA the main finding that COVID-19 deaths are associated with dissaving persists.

### Robustness checks
- Dependent variable denominator change:
  - Re-estimation with private savings as percent of GDP (instead of GPDI) — results:
    - Real per capita GDP growth remains positive and significant in several specifications.
    - COVID-19 cases per million negative and statistically significant when dependent variable = private savings/GDP (Column 5).
    - COVID-19 deaths per million negative and statistically significant when dependent variable = private savings/GDP (Column 9).
    - Stringency and vaccination remain non-significant in many robustness checks.
- Alternative proxy (bank deposits as share of monetary base):
  - Re-estimated from 2000 onward using central bank variables — GMM results largely inconclusive; COVID-19 variables not statistically significant in deposit-based regressions.
- Additional covariates tested (conflict dummy, informality, foreign aid, permanent/temporary GPDI components, uncertainty indices, etc.): none are statistically significant for SSA in reported GMM estimations, likely due to multicollinearity with baseline covariates.
- Overall robust conclusions from checks:
  - Real per capita GPDI growth is a robust positive determinant of private savings in SSA.
  - COVID-19 deaths per million are robustly associated with declines in private savings in SSA across multiple specifications and denominator choices.

### Policy implications and recommendations
- Raise real per capita economic growth:
  - Sustained real per capita GPDI growth is key to increasing private savings in SSA; policies and structural reforms to accelerate per capita growth are crucial.
- Health response to reduce mortality:
  - COVID-19 mortality had measurable adverse effects on private savings (30 deaths per million → ~0.6 percentage point decline); reducing mortality/severity through mass testing, vaccination rollout, and strengthened healthcare capacity can mitigate dissaving pressures.
- Social protection and targeted support:
  - Well-targeted social spending and poverty-reduction programs are needed to assist vulnerable households and prevent long-term depletion of savings, poverty increases, and adverse distributional effects.
- Fiscal policy interactions:
  - Public saving expansions show partial negative association with private savings (REH-consistent offset); fiscal policy choices should consider private saving dynamics.
- Continued data and vaccination efforts:
  - Low vaccination coverage through end-2021 (average ≈ 5 percent of population in SSA sample) limits detectability of vaccination effects; accelerating vaccination remains policy-relevant for both health and saving dynamics.

### Conclusions and directions for further research
- Contributions:
  - Revisits determinants of private savings in SSA; novel empirical investigation of COVID-19 impacts on private savings in SSA.
- Main takeaways:
  - Real per capita GPDI growth raises private saving rates in SSA (≈ 0.45–0.46 percentage points of GPDI per 1 percentage point growth).
  - COVID-19 deaths per million negatively associated with changes in private saving rates; every 10 deaths per million → ≈ -0.2 percentage points of GPDI change in private savings.
  - Microdata from household surveys corroborate macro and econometric evidence on savings depletion during COVID-19.
- Suggested further research:
  - Investigate association between private investment and private savings in SSA and COVID-19 effects on that dynamic.
  - Use additional microdata or higher-frequency macrodata (e.g., quarterly) to clarify transmission channels from COVID-19 to private savings and private investment.

*Source: wpiea2022176-print-pdf — IMF Working Paper "Private Savings and COVID-19 in Sub-Saharan Africa" (Introduction, Annex I, Results, and Conclusions).*

### Introduction............................................................................................................

### Introduction

### Background and objectives
- Low private saving rates in Sub-Saharan Africa (SSA) are a major development bottleneck, affecting consumption smoothing, risk management, and financing of life goals such as education and starting a business.
- Domestic savings are a key component for financing the sustainable development goals (SDGs) and for climate change mitigation.
- Prior to the pandemic, private saving rates in SSA had been catching up but remained below other emerging and developing economies (EMDEs) country groups, with significant heterogeneity across countries (notably low in fragile states and volatile in economies exposed to exogenous shocks).
- Access to formal financial products rose in recent years, but close to 46 percent of the population is still not able to save at all and less than a third of the existent savers in 2017 saved through formal financial institutions.
- The paper has two main objectives:
  - Reexamine the main determinants of private savings in the SSA region and compare them with other world regions.
  - Investigate the impact of COVID-19 (and associated preventive measures) on private savings in SSA—this is the first paper to do so for SSA according to the authors.

### Four analytical components performed in the paper
- Describe historical macro trends and stylized facts on SSA’s private savings at macro-level across sub-regions and country groups over the last four decades, including during COVID-19.
- Present stylized facts on the impact of COVID-19 on household savings at micro-level using UNDP household surveys from the second and third quarters of 2020 for selected SSA economies.
- Estimate empirically the determinants of private savings over the last four decades in SSA and compare with the full world sample and other world regions; baseline estimations are checked with robustness tests and inclusion of SSA-relevant variables (e.g., labor informality and years of conflicts).
- Econometrically investigate the effects of COVID-19 (cases and deaths per million people per country) on changes in private saving rates in SSA, controlling for other determinants and using the period sample 2017−21; also analyze the stringency of preventive measures and vaccination.

### Key empirical findings
- Real per capita economic growth is one of the most important determinants of private saving in SSA:
  - Every 1 percentage point increase in real per capita gross private disposable income (GPDI) growth (in PPP terms) is, on average, associated with an increase in 0.45 percentage points of GPDI in the SSA countries’ private savings rate.
  - The GDP-growth based specification yields robust results consistent with this finding.
- Other variables:
  - In baseline specifications, lagged dependent variable, rate of urbanization, and public savings are not significant; in some alternative estimations they are statistically significant with expected signs.
- COVID-19’s association with private savings:
  - The pandemic is, on average, negatively associated with the change in private savings of SSA households.
  - A statistically significant and negative coefficient is found for the number of COVID-19 deaths per country’s million inhabitants in estimations using first differences in private saving rates.
  - Every 10 COVID-19 deaths per million people in SSA countries is, on average, associated with a decline in 0.2 percentage points of GPDI in the change of their private savings.
  - The coefficient for cases per million people is also negative but not statistically significant in the authors’ estimations.
- Stringency and vaccination:
  - Tests for the stringency of preventive measures and for the number of vaccination shots as percentage of population produce inconclusive results; neither variable is statistically significant in estimations for SSA during the first two years of the pandemic.

### Micro and descriptive evidence
- Descriptive statistics and household surveys (UNDP, 2020) for selected SSA economies indicate declines in private savings during the pandemic, consistent with the econometric estimates.
- Other reports (WSBI, 2020; MasterCard Foundation, 2020) similarly suggest falls in private saving rates during COVID-19 in SSA.

### Policy implications and recommendations
- Boosting real per capita economic growth is critical to increase private saving rates in SSA; policies and structural reforms to recover from the pandemic and raise per capita growth are important.
- The negative association between COVID-19 deaths and changes in private saving rates underscores the need for a strong sanitary response, including mass testing and vaccination, to prevent continued adverse effects on private savings and the consequent risks to financing SDGs.

### Organization of the paper (as presented)
- Section 2: Descriptive statistics and stylized facts at macro level on private savings and COVID-19 across regions and SSA country groups.
- Section 3: Stylized facts from UNDP household surveys on micro-level impact.
- Section 4: Revisit main determinants of private savings in SSA and comparisons.
- Section 5: Investigate impact of COVID-19 (and preventive measures and vaccination) on private savings.
- Section 6: Robustness checks and additional tests.
- Section 7: Conclusions.

*Source: wpiea2022176-print-pdf - Introduction.*

### 17.3 percent in 2019 from 11.5 percent in 1983 (Figure 1, Panel 1).

### wpiea2022176-print-pdf - 17.3 percent in 2019 from 11.5 percent in 1983 (Figure 1, Panel 1)

### Regional trends and heterogeneity in private savings
- Private saving rate increased to 17.3 percent in 2019 from 11.5 percent in 1983 (Figure 1, Panel 1).
- Significant heterogeneity across SSA countries (Figure 2, Panel I).
- Oil exporters and middle-income countries (MICs) are the highest savers in the region (Figure 2, Panel II).
- Private saving rates are particularly low in fragile states and low-income countries (LICs).

### COVID-19 period: aggregate and comparative patterns
- In 2020 and 2021 (the COVID-19 period), private savings rates do not increase in SSA, in stark contrast with AEs (Figure 2, Panel II; and Figure 3, Panels I and II).
- For the sample of 36 SSA economies included in Figure 3, Panel I:
  - The average private saving rates in 2020 and 2021 remain at the same level as of 2019 (17.3 percent of GPDI).
  - The median private saving rates fell from 19.1 in 2019 to 18.3 percent of GPDI in 2020, reaching 18.8 percent of GPDI in 2021.
- Using the estimated sample of 31 SSA economies, the average decline in private savings combined in 2020 and 2021 is of -0.76 percent points of GPDI (see Table C in Annex II).
- By contrast:
  - In AEs the increase in private savings during the pandemic (2020 and 2021) is on average above 5 percent of GPDI.
  - In other EMDEs (excluding SSA), the average increase in private savings during 2020 and 2021 was of 1.9 percent of GPDI.
- For the full sample of LICs the average change in private savings for 2020 and 2021 is also negative.

### COVID-19 health data and underreporting considerations
- SSA economies have one of the lowest regional averages of officially reported cases and deaths per million people across the globe (Figure 4).
- Official statistics may be underestimated due to testing capacity, capacity to identify deaths from the disease, reticence of the local population to get tested or to go to hospital, and historical reasons.
- WHO (2021) estimates that, up to October 2021, COVID-19 cases and deaths in Africa could be seven times higher than the official statistics.
- In SSA there is no visible correlation between the number of cases and deaths per million (respectively) and the level or change in the savings rate (Figure 4, Panel I).

### Micro evidence: UNDP household surveys in six African economies (2020)
- Countries surveyed: Kenya, Nigeria, Rwanda, South Africa, Uganda, and Zambia.
- Survey sample sizes range from a minimum of 1,159 households for Zambia up to 6,120 households for Nigeria.
- Key survey findings (Figure 5 and Figure 6):
  - A significant share of African households has depleted its savings during COVID-19; savings depletion is the first or second source of financing during COVID-19 in all six countries investigated.
  - Savings are the main source of (emergency) income for households living both in rural and urban areas (Figure 6, Panel I), and for salaried professionals, self-employed and farming (Figure 6, Panel II).
  - A significant share of households reported that the depletion of savings is related to transportation issues and the possibility of supplying labor (most likely because of social distancing measures and lockdowns) during COVID-19 (Figure 6, Panel III).
  - Social distancing was the main form of COVID-19 prevention among households of all types of occupation (Figure 6, Panel IV).
- Selected precise shares from Figure 5 (Share of households surveyed, percent; the numbers on top of the bars represent the precise share):
  - Panel I - Kenya: Working 14.7, Savings 24.6, Loans 23.8, No source 36.8.
  - Panel II - Nigeria: Working 17.2, Savings 42.2, Loans 19.2, No source 21.4.
  - Panel III - Rwanda: Working 8.8, Savings 31.0, Loans 37.9, No source 22.3.
  - Panel IV - Uganda: Working 12.6, Savings 29.0, Loans 16.0, No source 42.5.
  - Panel V - South Africa: Working 10.9, Savings 22.0, Loans 18.8, No source 48.3.
  - Panel VI - Zambia: Working 32.7, Savings 24.1, Loans 28.8, No source 14.4.

### Interpretation of micro and macro evidence
- The household survey results corroborate the macro analysis: a significant share of households depleted savings during COVID-19, consistent with the observed lack of increase (and small decline) in aggregate private saving rates in SSA during 2020–2021.
- The survey was created for COVID-19; lack of comparable previous-year responses is a caveat for interpretation of the figures.

### Historical savings determinants and baseline econometric model
- The paper investigates historical determinants of private savings in SSA and compares results with the full world sample and different world income groups (AEs, EMs excluding SSA, LICs).
- Baseline panel estimation specification (model 1) used:
  - s_{i,t} = γ s_{i,t−1} + β B_{i,t} + δ X_{i,t} + α_i + τ_t + u_{i,t}, where:
    - s_{i,t} denotes private savings rates as percentage of GPDI for country i and time t;
    - B_{i,t} is a set of endogenous (and predetermined) covariates for savings;
    - X_{i,t} includes (strictly) exogenous variables;
    - α_i and τ_t are country- and time-fixed effects;
    - u_{i,t} are i.i.d error terms.
- Model (1) is estimated for an unbalanced panel data sample of at most 31 and at least 18 SSA economies.
- The section reports main results and results with additional determinants suggested by related literature.

*IMF Working Paper: Private Savings and COVID-19 in Sub-Saharan Africa*

### Annex I):  (i) the world sample, here containing 128 countries; (ii) LICs; (iii) EMDEs, excluding the SSA

### Annex I — Private Savings and COVID-19 in Sub-Saharan Africa (content unit)

### Econometric specification and estimator
- Baseline specification builds on Grigoli, Herman, and Schmidt-Hebbel (2018) with the real deposit rate and old dependency ratio excluded from baseline.
- Baseline regressors: lagged dependent variable; ln(real per capita GPDI in PPP); real growth rate of per capita GPDI in PPP; annual inflation rate; flow of private sector credit as percent of GPDI; public saving as percent of GPDI; terms of trade; share of urban population.
- First six explanatory variables assumed endogenous; last two treated as exogenous.
- Dynamic panel with small T and large N; primary preferred estimator: Two-stage System GMM (Blundell and Bond, 1998; Roodman, 2006). OLS with fixed effects (OLS FE) also reported for illustration.
- 2-Stage System GMM details: first differences and level equations used as instruments, robust standard errors, Windmeijer (2005) correction, Stata collapse to limit instrument proliferation. R-squared not reported for IV estimations; report F test, Hansen J-test, Arellano-Bond AR(1)/AR(2), and difference-in-Hansen.

### Data and sample construction
- Initial full sample: more than 133 countries, data 1980–2021 (published WEO and WDI April 2021 versions, augmented Grigoli et al. (2018) database).
- Outlier treatments: countries with population < one million in 2020 excluded; Zambia 2020–2021 excluded; CPI country-year observations above 95th percentile or below 5th percentile trimmed; real growth rate of per capita GPDI trimmed to below 99th and above 1st percentiles in some estimations.
- Final panel statistics highlighted in regressions:
  - SSA baseline: at most 31 countries, 986 observations in baseline regression.
  - Full (world) sample: 128 countries, 3,619 observations.
- COVID-19 variables: WHO-based cases and deaths from ourworldindata.org aggregated to annual 2020 and 2021, divided by country population to obtain per million metrics. Average (2020–2021) in the 13 SSA economies used for Equations (3) and (4): 4,190 cases per million; 59 deaths per million.
- Vaccination: aggregated total vaccine shots 2020–2021 divided by population; average vaccine shots in SSA sample by end-2021 ≈ 5 percent of population.

### Baseline empirical findings (historical determinants)
- Main robust historical determinant for SSA: real growth rate of per capita GPDI (PPP).
  - 2-Stage System GMM (SSA sample, Column (2) Table 1): a one percentage point increase in annual real per capita GPDI growth raises the private saving rate, on average, by 0.46 percentage points.
  - This 0.46 percentage points is noted as almost twice the average annual change in private savings rates in SSA over 1983−2019 (see Table C, Annex II).
- Full sample (world, 2-Stage System GMM): five variables statistically significant — lagged dependent variable; ln(real per capita GPDI); terms of trade; inflation; public saving/GPDI.
  - Public saving/GPDI coefficient (full sample) = -0.34 (Table 1), consistent with partial Ricardian offset.
- For LICs, EMDEs excluding SSA, and AEs: no statistically significant coefficients in baseline GMM estimations; GMM estimator appears restrictive and multicollinearity present.
- Interpretations noted:
  - Lagged dependent variable captures inertia in private savings.
  - Income growth (per capita GPDI) associated with higher private savings in SSA (supports Elbadawi and Mwega (2000) finding of causality from growth to savings in SSA).
  - Inflation ambiguous theoretically; in full sample inflation is negatively associated with private savings.
  - Terms-of-trade shocks can lead to higher private saving rates (positive effect in some samples).

### Tests of additional determinants (SSA, Eq. (2), 2-Stage System GMM)
- Additional covariates tested one-by-one (sample periods vary by variable availability): conflict dummy (UCDP), current account balance (% GPDI), real deposit rate, 5-year ahead forecast real GDP growth, foreign aid (% GPDI), labor informality (Medina and Schneider (2020) 1991–2017), permanent and temporary components of GPDI and TOT (HP filter λ = 6,25), old-age and young-age dependency ratios, unanticipated income growth (second difference of log income), unanticipated inflation (second difference of log CPI), Economic Uncertainty (World Uncertainty Index).
- Main result: none of the additional covariates are statistically significant for SSA in the GMM estimations reported (Tables 2a and 2b). Likely reason: multicollinearity with baseline covariates.
- Baseline variables that remain significant in some augmented regressions: real per capita GPDI growth; lagged dependent variable; ln(real per capita GPDI); flow of private sector credit/GPDI; share of urban population; public saving/GPDI.
  - Public saving/GPDI significant in multiple augmented regressions (Columns 1, 7, 15–17), corroborating REH effects in those settings.

### COVID-19 impacts on private savings — identification and specifications
- Two main first-difference specifications estimated to capture COVID-19 impact on year-to-year changes in private savings:
  - Equation (3): Δ(private savings rate) regressed on Δ lagged dependent variable, controls, and annual COVID-19 cases per million (country-year).
  - Equation (4): same as (3) but with annual COVID-19 deaths per million.
- COVID-19 variables considered strictly exogenous to private savings rates; caveat: misreporting biases likely downward, so estimates provide lower-bound effects.
- Period sample for COVID-19 impact estimations: 2017−2021 (deemed more relevant to capture pandemic years).

### COVID-19 estimation results — cases vs deaths
- COVID-19 cases per million:
  - Not statistically significant in 2-Stage System GMM estimations for SSA in Equation (3) (Table 3).
  - Some OLS FE specifications show negative associations but GMM (preferred) does not find significance for SSA sample.
- COVID-19 deaths per million:
  - Statistically significant and negative association with change in private savings rate in SSA in 2-Stage System GMM estimations (Table 4).
  - Economic magnitude example: a SSA country with 30 COVID-19 deaths per million in a year → predicted decline in private savings rate ≈ 0.6 percentage points of GPDI (ceteris paribus, average effect). This predicted change is close to the average first-difference of private savings between 2020 and 2021 (Table C, Annex II).
- Interpretation: COVID-19 mortality influenced agents toward dissaving in SSA; cases alone not statistically linked to decline in private savings in preferred estimations.

### Stringency of preventive measures and vaccination effects
- Stringency index (Oxford; Hale et al., ourworldindata.org), annual aggregation replacing COVID cases in Equation (3):
  - Not statistically significant for SSA or other samples in 2-Stage System GMM (Table 5). Conclusion: annual stringency measure does not capture forced-savings/foregone-consumption effects in sample.
- COVID-19 vaccination (total vaccine shots as percent of population, 2020–2021):
  - Not statistically significant in 2-Stage System GMM for SSA or other samples (Table 6).
  - Note: SSA average vaccine shots ≈ 5 percent of population by end-2021 in sample — low rollout may limit detectability of effects.
- Combined COVID-19 variables (cases/deaths with stringency or vaccination):
  - Combination improves statistical significance in some subsamples (Table 7):
    - For LICs (2-Stage System GMM): COVID-19 cases per million combined with vaccination rate → COVID cases coefficient negative and significant; vaccination shots coefficient positive and significant (suggests more vaccination associated with higher savings for LICs in that combination).
    - For EMDEs excluding SSA: same vaccination coefficient in that combination reported as statistically significant and negative (opposite effect).
  - For SSA, combined specifications do not overturn main finding that COVID-19 deaths are associated with dissaving (negative effect).

### Robustness checks
- Dependent variable denominator changes:
  - Re-estimated baseline and COVID-19 specifications using private savings as percent of GDP (instead of GPDI) and mixed specifications (explanatory vars still percent of GPDI).
  - Key robustness outcomes (Table 8):
    - Real per capita GDP growth remains statistically significant and positive in several specifications.
    - COVID-19 cases per million negative and statistically significant when dependent variable = private savings/GDP (Column 5).
    - COVID-19 deaths per million negative and statistically significant when dependent variable = private savings/GDP (Column 9). Confirms robustness of negative deaths effect.
    - Stringency and vaccination remain non-significant in many robustness checks.
- Alternative dependent variable: bank deposits as share of monetary base (proxy for private savings), using IFS central bank survey variables (liabilities to other depository corporations / monetary base) — data available from 2000 onward.
  - Re-estimated Equations (1)–(4) and variants starting 2000 sample; panel and sample size changes (Table 9).
  - Results: many coefficients similar in sign but GMM estimations with deposits over monetary base produce largely inconclusive results — few statistically significant coefficients; COVID-19 variables not statistically significant in these deposit-based regressions.
  - Conclusion of robustness checks: primary finding (real per capita GPDI growth positive determinant; COVID-19 deaths negatively associated with private saving rate) robust across denominator change to GDP; alternative bank-deposit proxy yields inconclusive GMM results.

### Policy-relevant implications and takeaways
- Historical policy implication:
  - Sustained real per capita GPDI growth is a key driver of higher private savings in SSA; policies that raise per capita income growth can support higher private saving rates.
  - Public saving expansions (fiscal consolidation) are associated with lower private savings (REH-consistent negative coefficient), but offset is partial (e.g., full-sample public saving/GPDI coefficient = -0.34).
- COVID-19 response implications:
  - COVID-19 mortality had a measurable adverse effect on private savings in SSA (30 deaths per million → ~0.6 percentage point decline in private savings rate). Thus, measures that reduce mortality/severity (vaccination rollout, healthcare capacity) can mitigate dissaving pressure.
  - Stringency of preventive measures and vaccination rates (as observed through 2021) do not show consistent significant effects on private savings in SSA in preferred GMM estimations — low vaccine coverage and data limitations may underlie these null findings.
  - Continued acceleration of pandemic response in SSA, particularly vaccination, is important to reduce long-term risks and to support private saving dynamics during recovery.

*Italic: Source — IMF Working Paper ("Private Savings and COVID-19 in Sub-Saharan Africa"), Annex I and related tables and text (content unit: wpiea2022176-print-pdf - Annex I))*

### Conclusions

### Conclusions

### Contributions and research questions
- Revisits the main determinants of private savings in the SSA region and compares them with other country groups.
- First paper, to the authors' knowledge, to investigate the impact of COVID-19 on private savings in SSA.

### Methods and analytical approach
- Four types of analyses:
  - Description and analysis of historical macro trends and stylized facts on SSA’s savings across sub-regions and country groups over the last four decades, including macro-level effects of COVID-19 on private savings in 2020 and 2021.
  - Micro-level stylized facts from household surveys prepared by the United Nations Development Program (UNDP, 2020) in Q2 and Q3 2020 in selected SSA economies.
  - Empirical estimation of determinants of private savings in SSA over four decades and comparison with other world regions.
  - Econometric investigation of the effects of COVID-19 (cases and deaths) on the change in saving rates in SSA, controlling for other determinants; additional focus on stringency of preventive measures and COVID-19 vaccination.
- Econometric specification follows Grigoli, Herman, and Schmidt-Hebbel (2018) with additional tests, including on labor informality.

### Key empirical findings
- Real per capita economic growth:
  - Every 1 percentage point increase in real per capita GPDI growth is, on average, associated with an increase in approximately 0.45 percentage points of GPDI in the SSA countries’ private savings rate.
  - An increase of 2 percentage points (around the sample average of real per capita GPDI growth between 2020 and 2021 in SSA), ceteris paribus, would be expected to increase the private savings rate in SSA by 0.9 percentage points of GPDI.
- COVID-19 mortality effects:
  - Private saving rates have, on average, not increased or even marginally declined in SSA during the pandemic period.
  - The number of COVID-19 deaths per million people is statistically significant and negatively associated with the change in private savings in SSA.
  - Every 10 cumulative deaths of COVID-19 per million inhabitants (around the median between 2020−21 in SSA) is, ceteris paribus and on average, associated with a decrease of around 0.2 percentage points of GPDI in the change of the private savings rate.
  - Using 59 deaths per million inhabitants (approximately the average between 2020−21 in SSA) implies, ceteris paribus, a decline of close to 1.2 percentage points of GDPI in the change of SSA’s private savings rate.
- Microdata corroboration:
  - Macro-level descriptive statistics and econometric results are corroborated by micro analysis of household surveys in selected SSA economies (see WBSI and MasterCard Foundation, 2020), which point to the harmful effect of COVID-19 at the household level.

### Comparative and welfare implications
- Findings for SSA contrast with those for Advanced Economies (AEs) (e.g., McGregor, Suphaphiphat, and Toscani, 2022).
- Private savings are a key source of financing for post-pandemic relaunch and movement toward the SDGs (Gaspar et al., 2019; and Benedek et al., 2021).
- Depletion of household savings and increases in poverty should be addressed to avoid longer-term negative impacts, including on income inequality and development (Agarwal and others, 2022).
- Multilateral institutions and development agencies have called for substantial and well-targeted social and financial supports from African governments to the most vulnerable populations (see UN, 2020; and IMF, 2021a).

### Policy recommendations
- Continue adopting policies and structural reforms to accelerate real per capita economic growth in the region.
- Build institutions to spur spillover effects from external demand and mitigate negative external shocks.
- Support initiatives to reduce the spread of COVID-19 in the SSA region.
- Implement well-targeted social spending and poverty-reduction programs to assist households during the economic recovery period.
- Provide social and financial supports targeted to the most vulnerable populations.

### Directions for further research
- Investigate the association between private investment and private savings in SSA and how COVID-19 may have affected that dynamics.
- Conduct more detailed analysis of COVID-19 impacts on private savings using further microdata or higher-frequency macrodata (e.g., quarterly) to shed light on transmission channels into private savings and private investment in SSA.

*Source: Conclusions, wpiea2022176-print-pdf - Conclusions*

### References

### References

### Savings, Terms-of-Trade, and Macroeconomic Dynamics
- Adler, Gustavo, and Nicolas E. Magud, 2013. “Four Decades of Term s-of-Trade Booms: Saving-Investment Patterns and a New Metric of Income Windfall.” IMF Working Paper No. 13/103, May.
- Agénor, Pierre-Richard, and Joshua Aizenman, 2004. “Savings and the terms of trade under borrowing constraints.” Journal of International Economics, Vol. 63, pp. 321−340.
- Grigoli, Francesco, Alexander Herman, Klaus Schmidt-Hebbel, 2018. “Saving in the World” World Development, Vol. 104, pp. 257−270.
- Elbadawi, Ibrahim A., and Francis M. Mwega, 2000. “Can Africa's Saving Collapse Be Reversed?” World Bank Economic Review, Vol. 14, No. 3, pp. 415−443.
- Schclarek, Alfredo, and Mauricio Caggia, 2015. “Household Saving and Labor Informality: The Case of Chile” IDB Working Paper, No. IDB-WP-581, May.
- Shawa, Ken Chamuva, 2016. “Drivers of Private Saving in Sub-Saharan African Countries,” Journal of Economic Development, Vol. 41, No. 2, pp. 77−110.
- Gruss, Bertrant, Malhar Nabar, and Marcos Poplawski-Ribeiro, 2020. “Growth Accelerations and Reversals in Developing Economies: External Conditions and Domestic Amplifiers,” Open Economies Review, Vol. 31, pp. 753−786.
- Dezso, Diana, James Robinson, and Nisha Sing, 2018. “Savings count: Trends in access, use and the ecosystem of savings in Sub-Saharan Africa,” Itad Report (Hove: Mastercard Foundation Savings Learning Lab).
- Dobson, Stephen, Carlyn Ramlogan-Dobson, and Eric Strobl, 2020. “Savings and the Informal Sector,” International Review of Applied Economics, Vol. 34(2), pp. 1−18.
- Buffie, Edward F., Christopher Adam, Luis-Felipe Zanna, and Kangni Kpodar, 2022. “Loss-of-Learning and the Post-Recovery in Low-Income Countries.” IMF Working Paper, No. 22/25, February.
- Attinasi, Maria Grazia, Alina Bobasu and Ana-Simona Manu, 2021. “The implications of savings accumulated during the pandemic for the global economic outlook.” ECB Economic Bulletin, Issue 5/2021.
- McGregor, Thomas, Nujin Suphaphiphat, and Frederik Toscani, 2022. “Excess Savings in the Euro Area.” IMF Working Paper, forthcoming.

### COVID-19, Health Outcomes, Vaccination, and Mortality Measurement
- Agarwal, Ruchir, and Gita Gopinath, 2021. “A Proposal to End the COVID-19 Pandemic.” IMF Staff Discussion Note, No. 2021/004, May.
- Agarwal, Ruchir, Jeremy Farrar, Gita Gopinath, Richard Hatchett, and Peter Sands, 2022. “A Global Strategy to Manage the Long-Term Risks of COVID-19.” IMF Working Paper, No. 22/68, April.
- Agrawal, Utkarsh, Srinivasa Vittal Katikireddi, Colin McCowan, Rachel H Mulholland, Amaya Azcoaga-Lorenzo, Sarah Amele, Adeniyi Francis Fagbamigbe, Eleftheria Vasileiou, Zoe Grange, Ting Shi, Steven Kerr, Emily Moore, Josephine L K Murray, Syed Ahmar Shah, Lewis Ritchie, Dermot O’Reilly, Sarah J Stock, Jillian Beggs, Antony Chuter, Fatemah Torabi, Ashley Akbari, Stuart Bedston, Jim McMenamin, Rachael Wood, Ruby S M Tang, Simon de Lusignan, F D Richard Hobbs, Mark Woolhouse, Colin R. Simpson, Chris Robertson, Aziz Sheikh, 2021. “COVID-19 hospital admissions and deaths after BNT162b2 and ChAdOx1 nCoV-19 vaccinations in 2·57 million people in Scotland (EAVE II): a prospective cohort study.” The Lancet Respiratory Medicine, Vol. 9, Issue 12, pp.1439–1449.
- Aizenman, Joshua, Alex Cukierman, Yothin Jinjarak, Sameer Nair-Desai, and Weining Xin, 2022. “A Two Covid-19 Years Quartile Comparison of Official with Excess Mortality: Voice and Accountability and the Impact of Vaccines.” NBER Working Paper, No. 29778, February.
- Karlinsky, Ariel, and Dmitry Kobak, 2021. “Tracking excess mortality across countries during the COVID-19 pandemic with the World Mortality Dataset.” eLife Vol. 10, pp. e69336. DOI: https://doi.org/10.7554/eLife.69336.
- World Health Organization (WHO), 2021. “Six in seven COVID-19 infections go undetected in Africa,” WHO Regional Office for Africa Statement (available at https://www.afro.who.int/news/six-seven-co vid-19-infections-go-undetected-africa).
- Ritchie, Hannah, Edouard Mathieu, Lucas Rodés-Guirao, Cameron Appel, Charlie Giattino, Esteban Ortiz-Ospina, Joe Hasell, Bobbie Macdonald, Diana Beltekian and Max Roser, 2020. “Coronavirus Pandemic (COVID-19).” Published online at OurWorldInData.org. Retrieved from: 'https://ourworldindata.org/coronavirus'
- Hakobyan, Shushanik, 2021. “In the Race to Vaccinate Sub-Saharan Africa Continues to Fall Behind.” IMF Blog, November 22nd.
- Nyantakyi, Eugene Bempong, and Jonathan Munemo, 2021. “Going the Last Mile. Improving Sub-Saharan Africa’s logistics could be the key to successful vaccine delivery.” Finance & Development, December, pp. 17−19.
- Miguel, Edward, and Ahmed Mushfiq Mobarak, 2021. “The Economics of the COVID-19 Pandemic in Poor Countries.” NBER Working Paper, No. 29339, October.
- Ercolani, Valerio, Elisa Guglielminetti, and Concetta Rondinelli, 2021. “Fears for the future: saving dynamics after the Covid-19 outbreak”, Bank of Italy Covid-19 Note, June.
- Lee Smith, A., 2020. “Why Are Americans Saving So Much of Their Income?” Federal Reserve Bank of Kansas City Economic Bulletin, December, pp.[...]
- World Savings and Retail Banking Institute and MasterCard Foundation, 2020. “Savings and Retail Banking in Africa: A case study on Covid-19: from a customer, retail banking and regulatory perspective” Scale2Save Report, November (available at https://www.wsbi-esbg.org/SiteCollectionDocuments/1354_ESBG_BRO_SCALE2SAVE_EN_FINAL_DEC_2020.pdf)
- Wiley Population Council, 2015. “Socioeconomic Effects of Ebola in Africa.” Population and Development Review, vol. 41, no. 2, pp. 372–75.

### Sub-Saharan Africa: Fiscal Policy, Development, and Structural Challenges
- International Monetary Fund (IMF), 2018a. Chapter 2: “Domestic Revenue Mobilization in Sub-Saharan Africa: What Are the Possibilities?” in African Regional Economic Outlook, April (Washington DC: International Monetary Fund).
- International Monetary Fund (IMF), 2018b. Chapter 2: “Capital Flows to Sub-Saharan Africa: Causes and Consequences,” in African Regional Economic Outlook, October (Washington DC: International Monetary Fund).
- International Monetary Fund (IMF), 2019. Chapter 2: “Domestic Arrears in Sub-Saharan Africa: Size, Effects and Solutions” in African Regional Economic Outlook, October (Washington DC: International Monetary Fund).
- International Monetary Fund (IMF), 2020. “A Difficult Road to Recovery” in African Regional Economic Outlook, October (Washington DC: International Monetary Fund).
- International Monetary Fund (IMF), 2021a. “.” IMF Sub-Saharan African Regional Economic Outlook, October 2021.
- International Monetary Fund (IMF), 2021b. “Divergent Recoveries and Global Imbalances.” IMF External Sector Report 2021.
- Gaspar, Vitor, David Amaglobeli, Mercedes Garcia-Escribano, Delphine Prady, and Mauricio Soto, 2019. “Fiscal Policy and Development: Human, Social, and Physical Investment for the SDGs” IMF Staff Discussion Note 19/03 (Washington DC: International Monetary Fund).
- Lledó, Victor, and Marcos Poplawski-Ribeiro, 2013. “Fiscal Policy Implementation in Sub-Saharan Africa,” World Development, Vol. 46(C), pp. 79–91.
- World Bank, 2022. “Tanzania Economic Update— Empowering Women: Expanding Access to Land, Assets and Economic Opportunities” World Bank Report, January.
- Selassie, Abebe Aemro, 2018. “The Debt Challenge to African Growth,” Op-ed (https://www.imf.org/en/News/Articles/2018/07/31/vc052318-the-de bt-challenge-to-african-growth) issued on May 23rd (Washington: International Monetary Fund).
- Belianska, Anna, Nadja Bohme, Kaihao Cai, Yoro Diallo, Saanya Jain, Giovanni Melina, Marcos Poplawski-Ribeiro, and Solo Zerbo, 2022. “Climate Change and New Financial Instruments: Opportunities and Challenges for Sub-Saharan Africa,” IMF Staff Climate Notes, forthcoming.

### Conflict, Political Economy, and Social Impacts on Savings and Growth
- Dube, Oeindrila, and Juan F. Vargas, 2013. “Commodity Price Shocks and Civil Conflict: Evidence from Colombia.” Review of Economic Studies, Vol. 80, No. 4, pp. 1384−1421.
- Torres-Garcia, Alejandro, Martin Vanegas-Arias and Laura Builes-Aristizabal, 2019. “The Effect of Armed Conflict on Savings Rates: International Evidence from 1980 to 2015.” Peace Economics, Peace Science and Public Policy, Vol. 25, No. 3, pp. 20180048.
- Voors, Maarten J., Eleonora EM Nillesen, Philip Verwimp, Erwin H Bulte, Robert Lensink, Daan P Van Soest, 2012. “Violent conflict and behavior: a field experiment in Burundi” American Economic Review, Vol. 102(2), pp. 941−964.
- Novta, Natalija, and Evgenia Pugacheva, 2021. “The macroeconomic costs of conflict.” Journal of Macroeconomics, Vol. 68, No. 103286.
- Wiley Population Council, 2015. “Socioeconomic Effects of Ebola in Africa.” Population and Development Review, vol. 41, no. 2, pp. 372–75.

### Econometric Methods, Data Sources, and Uncertainty Measures
- Arellano, Manuel, and Olympia Bover, 1995. “Another look at the instrumental variable estimation of error-components models,” Journal of Econometrics, Vol. 68(1), pp. 29–51.
- Blundell, Richard W., and Stephen R. Bond, 1998. “Initial conditions and moment restrictions in dynamic panel data models.” Journal of Econometrics, Vol. 87(1), pp. 115–143.
- Bond, Stephen, Anke Hoeffler, and Jonathan Temple, 2001. “GMM Estimation of Empirical Growth Models,” Economics Papers 2001-W21, Nuffield College, University of Oxford.
- Baum, Christopher F., Mark Edwin Schaffer, and Steven Stillman, 2003. “Instrumental variables and GMM: Estimation and testing.” The Stata Journal, Vol. 3(1), pp. 1–31.
- Roodman, David, 2006. “How to do xtabond2: An introduction to ‘difference’ and ‘system’ GMM in Stata.” Center for global development working paper no. 13. Washington, December.
- Roodman, David, 2007. “A note on the theme of too many instruments.” Center for global development working paper no. 125. Washington, December.
- Hausman, Jerry, 2001. “Mismeasured Variables in Econometric Analysis: Problems from the Right and Problems from the Left,” Journal of Economic Perspectives, Vol. 15, No. 4, pp. 57−67.
- Arellano, Manuel, and Olympia Bover, 1995. “Another look at the instrumental variable estimation of error-components models,” Journal of Econometrics, Vol. 68(1), pp. 29–51.
- Ahir, Hites, Nicholas Bloom, and Davide Furceri, 2022. “World Uncertainty Index.” NBER Working Paper, No. 29763, February.
- Aizenman, Joshua, Eduardo Cavallo, Ilan Noy, 2015. “Precautionary Strategies and Household Saving.” NBER Working Paper, No. 21019, March.

*Private Savings and COVID-19 in Sub-Saharan Africa — Working Paper No. WP/2022/176*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022176-print-pdf.pdf_