Geospatial patterns and socio-environmental factors of household overcrowding in Ethiopia: Evidence from 2019 Ethiopian demographic and health survey data – Nature. Com 09:55

Geospatial patterns and socio-environmental factors of household overcrowding in Ethiopia: Evidence from 2019 Ethiopian demographic and health survey data

Scientific Reports volume 16, Article number: 8504 (2026) Cite this article

Abstract

United Nation Sustainable Development Goals (SDG) Goal 11 aims make cities and human settlements safe, resilient, inclusive, and sustainable, with special emphasis on reducing overcrowding, adequate and healthy housing. However, many governments in the African continent are facing challenges with inadequate housing as population growth and livelihood coupled with insufficient infrastructure, grow faster than having affordable housing. This research study sought to explore spatial differences and factors associated with overcrowding among households in Ethiopia. A community-based cross-sectional study design was utilized to engage a nationally representative dataset to understand the demographic, socio-economic and environmental factors of household overcrowding. Data analysis employed STATA 17, ArcGIS 10.8, and Kulldorff’s Sat Scan 10.2.5, with sampling weights applied to ensure robust statistical estimates. The prevalence of household overcrowding in Ethiopia was (69.62%), showing a higher burden in the Somali region (78.8%) and lower burden in Addis Ababa City (47.3%). The Sat Scan analysis revealed a significant cluster in the south and central regions (LLR = 9.10, P = 0.0051). Geographically weighted regression analysis revealed that lack of education, no attainment of primary education, lack of media exposure, more than five family members, and use of pastoralist livelihoods were spatial factors of household overcrowding in Ethiopia. The prevalence of household overcrowding in Ethiopia shows a relatively high prevalence compared to other countries in Sub-Saharan Africa. This study also showed that lack of education, limited attainment at the primary level, no media exposure, more than five family members, living on pastoralist region significant spatial factors of household overcrowding. Geographically tailored interventions including access to information in the northwest while higher education, improved livelihoods, infrastructure development, and culturally appropriate housing solutions in the east are required.

Introduction

Adequate indoor space means having enough room in a house to prevent overcrowding and components of healthful housing. It helps to maintain clean air, prevent the spread of infectious diseases, protect from exposure to harmful noise, provide privacy, sense of security and stability, and satisfy other basic requirements that support physical and mental health1,2. However, people and families with overcrowded housing may not have enough room to sleep, cook, study, play, and have privacy. Household overcrowding (HC) reduces comfort, negatively affects quality of life and it is linked to an increased risk of infectious disease, increased stress and anxiety, disrupted child development, family conflict, and household safety. It is a measure of poor socioeconomic status, social deprivation, and a stressful social environment that leads to high mortality3,4,5,6.

In contemporary public health discourse, sustainable housing is increasingly recognized as a critical determinant of health, functioning as a setting where interdependent physical, social, economic, behavioral and environmental factors converge and interact in complex ways to shape the well-being and health outcomes of occupants5,6,7. The World Health Organization (WHO) housing and health guidelines show that household overcrowding is a global concern. It is defined as the health and social risks associated with insufficient living, sleeping, and household space for the number of occupants in a dwelling8.

Wide range of infectious diseases such as tuberculosis9,10, hospitalizations due to influenza11,12, pneumonia, acute respiratory infections12,13, meningococcal disease14,15, rheumatic fever, skin infections16,17, and intestinal diseases18,19 has been associated with HC. It has linked with non-communicable diseases and psychological problems20. Household overcrowding also affects various aspects of the lives of children, including their academic performance, the social relationships among women and children, the care given to children21,22 and mental health23,24.

The definitions of household overcrowding vary across institutions and countries. The U.S. Department of Housing and Urban Development commonly defines overcrowding as more than one person per room and severe overcrowding as more than 1.5 persons per sleeping room25 while the WHO considers average living area per person8. For this study, United nation (UN)-Habitat definition was used, HC occurring when more than two persons in a family share a single sleeping room (excluding bathrooms but including kitchens and living rooms)26.

The global population has significantly grown over the last century and is projected to reach 9.7 billion by 2050, with developing regions, particularly Africa, expected to bear the greatest demographic pressures27,28. Overcrowding is widely prevalent in many low- and middle-income countries that is expected to worsen and intensify with rapid urbanization. A study conducted in China in 2016 reported that 47.5% of urban households experienced overcrowding, which is defined as more than two people per habitable room29. Similarly, research conducted in Nigeria in 2015 revealed that households of 8 to 12 individuals were commonly accommodated in dwellings containing only two rooms30. In Ghana, an estimated 44.5% of households are affected by overcrowded living conditions31. Around 57.6% of dwellings in the suburb of Johannesburg, South Africa were overcrowded26.

Ethiopia is experiencing rapid population growth and urbanization leading to overcrowding and poor housing conditions especially in Addis Ababa where most household (67.42%) lived in overcrowded housing32. Houses in urban areas can be overcrowded because of rapid migration, rampant slums, and the high housing costs, whereas rural households may be constrained by poverty, inadequate housing quality, and cultural norms. Environmental stressors including drought, conflict induced displacement, and climate mediated migration further stress household accommodation leading to spatial hotspots of vulnerability that are not well mapped. Moreover, most urban housing is informal, substandard, and in need of replacement, while large-scale initiatives such as the Integrated Housing Development Program have fallen far short of addressing the growing housing deficit33. These conditions indicate overcrowding as a pressing public health and social challenge in Ethiopia thereby justifying the need for research and policy attention.

Previous studies conducted on household crowding in Ethiopia is predominantly localized and descriptive in nature despite its enormity and lacks investigations on spatial heterogeneity and the social-environmental factors influencing the variation in household crowding. Classical regression models assume same factors across space in order for their estimates to be efficient and statistically consistent but the reality on the ground is such that the factors of HC crowding including socioeconomic and demographic, housing and environmental factors across the country. Understanding these spatial variations is important, as national averages hide relevant local variations and the need for geographically targeted measures.

Thus, this method will be able to identify areas of significant HC and investigating the association between social and environmental factors in spatial context. This evidence informs policymakers and decision-makers to take necessary actions to achieve Sustainable Development Goals, particularly SDG 3 (Good Health and Well-being), SDG 10 (Reduced Inequalities), and SDG 11 (Sustainable Cities and Communities) in Ethiopia. It is also crucial in guiding housing and public health interventions. This study showed the detection of HC as not only indicator of poor socioeconomic status but also geographically patterned, multidimensional public health and development problem, which requires spatial-based responses.

Methods and materials

Study design and setting

This study was a community based cross sectional study utilizing secondary data available from 2019 Ethiopia Mini Demographic and Health Survey (EMDHS) conducted Nationwide. Ethiopia is situated in Northeast Africa, between 3° and 15° north latitude and 33° and 48° east longitude, with an area of about 1.13 million km². Administratively, Ethiopia is made up of eleven regional states:1) Tigray, 2) Afar, 3) Amhara, 4) Oromia, 5) Somali, 6) Benishangul-Gumuz, 7) South West Ethiopia, 8) Sidama, 9) Southern Nations, Nationalities and Peoples’ Region [SNNP], 10) Gambella, and 11) Harari, with two chartered city administrations (Addis Ababa and Dire Dawa). Regions are further divided into zones, districts, and kebeles (the smallest administrative unit)34.

Sampling procedures

To guarantee representativeness at the national, regional, and urban–rural levels, the 2019 EMDHS used a two-stage stratified cluster sampling design. Using a probability proportional to size based on the 2007 Population and Housing Census sampling frame, 305 Enumeration Areas35 were chosen in the first stage, comprising 93 urban clusters and 212 rural clusters. A nationally representative 8,663 households were obtained in the second stage by systematically selecting a fixed number of households from each EA.

For this study, the household (HR) dataset was extracted since it provides detailed information on household members, dwelling characteristics, and housing conditions necessary for assessing household overcrowding. Main variables included household size, number of rooms used for sleeping, socioeconomic indicators, and housing factors. To maintain representativeness and account for sampling probabilities, the sampling weights provided in the 2019 EMDHS dataset were applied in all analyses34.

Study population and sample size

The study population included 8663 households surveyed in the 2019 EMDHS. Household overcrowding was defined as the number of household members divided by the room used for sleeping, according to standard definitions from the United Nations Habitat and WHO housing guidelines8,26.

Study variables

Dependent variable

Household overcrowding was defined at the household level as a binary variable (either overcrowded or not overcrowded), indicating whether the number of usual household members exceeded the number of sleeping rooms. It was defined using the persons per sleeping room (PPR) measure, which is calculated by dividing the number of usual household members (de jure members that is persons who usually live in the household) by the number of sleeping rooms. A threshold of two persons per sleeping room (PPR ≥ 2) was used to classify households as overcrowded, consistent with DHS methodology36 and similar to the two-persons-per-bedroom standard used by U.S. Department of Housing and Urban Development25. For spatial analyses and geographically weighted regression (GWR), household-level data were aggregated to the survey cluster level by calculating a weighted proportion of overcrowded households using DHS sampling weights.

The cluster-level overcrowding proportion was calculated as:

where denotes household overcrowding status (1 = overcrowded, 0 = not overcrowded) and represents the DHS sampling weight.

Independent variables

Independent variables were selected based on empirical relevance and previous research evidence. They were classified as socioeconomic and demographic, housing, and tenure factors. Socio-demographic factors included the age of the household head (in years), sex of the household head (male/female), type of residence (urban/rural), region of region (agrarian, pastoralist, city), household size (number of usual members), educational status (no education, primary, secondary, or higher) and media exposure (yes/no). Socioeconomic factor was as wealth index (poorest, poorer, middle, richer, and richest). The housing characteristics included tenure status, number of rooms in the dwelling used for sleeping, type of sanitation services (basic vs. not basic), and housing material (floor, wall, and roof type).

A household had access to basic sanitation if it had an unshared flush or pour-flush toilet connected to a piped sewer system, a septic tank, or pit latrines, or if it used a ventilated improved pit latrine, composting toilet, or pit latrine with a slab and did not know where it disposed of its excreta37. Media exposure was calculated using three indicators from the 2019 EMDHS dataset: whether a household respondent reported reading newspapers or magazines, listening to the radio, or watching television. A binary variable was created by combining these indicators, such that households in which the respondent was exposed to at least one of the three media sources were categorized as “yes” (coded = 1), while those with no exposure to any of the sources were classified as “no” (coded = 0)34.

Improved floor material was computed as a household having a floor made of cement, tiles, parquet/polished wood, or other finished floor material. Improved wall material was defined as a household having walls made of cement block, bricks, stone with lime/cement, or other finished wall material, and improved roof material represented household having a roof made of corrugated iron, roof tiles, concrete, or other finished roof material38,39.

Data management

The data were properly managed using STATA/SE version 17.0 and Microsoft Excel version 16, which included editing, verifying, arranging, cleaning, and recoding. The proportions of each dependent and independent variable were calculated using cross-tabulation with the variable cluster number (hv001) and saved to Excel as a CSV file. Variables with a linear relationship to the outcome variable were imported into ArcGIS 10.8 for the Ordinary Least Squares40 model. To account for the effect of the complex sampling design of the survey and the hierarchical nature of the EDHS dataset, the data were weighted using the “svyset” command in STATA. This command was used as a prefix for all analyses in this study to ensure representativeness and generate reliable statistical estimates.

All household-level analyses were performed using survey weights to account for the complex multistage sampling design of the DHS. The household weight variable (hv005) was divided by 1,000,000 as recommended in the DHS Guide. The unweighted sample included 8,663 households, which after applying weights represent approximately 22.4 million households nationally. Descriptive statistics (percentages and means) are reported as weighted estimates, while sample counts (n) are unweighted.

Statistical analysis

All descriptive statistical analyses incorporated DHS survey weights, strata, and primary sampling units to account for the complex survey design. Spatial analyses, including hotspot analysis (Gi*), kriging, Sat Scan, and geographically weighted regression (GWR), utilized cluster-level weighted proportions to approximate population patterns. Due to software limitations, these spatial methods did not fully account for the survey design features; therefore, results should be interpreted with caution in areas with few clusters or highly variable sampling weights. As DHS cluster coordinates are randomly displaced to protect respondent confidentiality, spatial analyses were interpreted at an aggregated scale (cluster and administrative-unit level) rather than fine-scale geographic inferences (exact point locations).

Spatial autocorrelation and hotspot analysis

The presence of spatial variation and the mapping of model parameters between local models were investigated using ArcGIS version 10.8 software. The global spatial autocorrelation (Global Moran’s I) was used to determine whether household overcrowding was dispersed, clustered, or randomly distributed in Ethiopia. Global Moran’s I is a spatial statistic that measures spatial autocorrelation across a dataset and returns a single output value ranging from − 1 to + 1. A statistically significant Moran’s I (p < 0.05) suggests that household overcrowding is not distributed at random (clustered or dispersed). Hotspot analysis was conducted using the Getis-Ord Gi statistic to identify spatial clustering of high and low values of household overcrowding. Getis-Ord Gi* was selected over Local Moran’s I because the primary objective was to detect broader geographic concentrations (hotspots and cold spots) rather than local spatial outliers. Spatial relationships were defined using a fixed distance-band spatial weights matrix, with the distance threshold selected based on the average inter-cluster spacing to ensure each cluster had a sufficient number of neighbors. Sensitivity analyses using alternative distance thresholds produced consistent hotspot patterns. Gi* Z-scores and associated p-values were interpreted descriptively, consistent with exploratory spatial data analysis practice, and no formal multiple-testing correction was applied.

Spatial interpolation

The ordinary Kriging spatial interpolation method was used to predict the proportion of household overcrowding in unsampled locations using values from neighboring clusters. Ordinary Kriging was selected over other interpolation techniques because it is an optimal interpolator that provides the smallest mean error (ME) and root mean square error (RMSE) when estimating spatial patterns [39]. It was applied as a descriptive spatial smoothing technique using default parameters to visualize geographic variation in household overcrowding. The kriged surfaces are intended for visualization purposes only and are not interpreted as optimal predictive estimates.

Spatial scan analysis

Kulldorff’s purely spatial scan statistics with a Bernoulli model were employed to detect and locate statistically significant clusters of household overcrowding using Sat Scan™ version 10.2.5 software. For spatial cluster detection using SaTScan, a Bernoulli probability model was applied, where overcrowded households were treated as cases and non-overcrowded households as controls. The case file, control file, and coordinate file (latitude and longitude) were imported into Sat Scan™ to identify the most likely clusters. The maximum scanning window size was set as a proportion of the total population at risk, with the upper limit restricted to less than 50% to avoid overlooking both very small and very large clusters. Primary (most likely) and secondary clusters were identified using p-values and likelihood ratio tests, with the cluster showing the maximum likelihood ratio designated as the most likely cluster. Statistical significance of spatial clusters was evaluated using 999 Monte Carlo replications. Relative risks (RRs) and corresponding 95% confidence intervals were estimated for statistically significant clusters.

Factors associated with household overcrowding

The Ordinary Least Squares40 regression model was first fitted as a global model to estimate the relationship between household overcrowding and the independent variables. OLS assumes that the effect of factors remains constant throughout the study area. OLS assumes that the effect of factors remains constant across the entire study area. This step was important for selecting the appropriate independent variables associated with spatial variation in overcrowding. Prior to fitting the global and local models, the assumption of spatial stationarity was tested using global spatial autocorrelation. Afterward, the global spatial regression model was calibrated to identify key factors associated with the prevalence of household overcrowding.

The six assumptions of the OLS model were evaluated before advancing to geographically weighted regression (GWR). These included: (1) ensuring explanatory variables had the expected direction of association, (2) statistical significance of each variable, (3) randomness of residuals, (4) non-significance of Jarque-Bera statistics for normality, (5) acceptable Variance Inflation Factor (VIF) values for multicollinearity, and (6) adequate explanatory power as measured by R-squared. Multicollinearity was specifically assessed using VIF values to confirm the robustness of the model. Factors with Variance Inflation Factor (VIF) values greater than 7.5 were considered indicative of multicollinearity. In this study, all factors had VIF values below the threshold, suggesting that no evidence of multicollinearity was observed in the data.

After checking the assumptions of the OLS model, Geographically Weighted Regression (GWR) was employed to capture spatially varying relationships, which assume that associations between variables may differ across locations [43]. In this study, the analysis was conducted using ArcGIS version 10.8 software. For model comparison, AICc and Adjusted R² were used as selection criteria between the global regression40 and local regression (GWR), with the model yielding the lowest AIC and higher Adjusted R² considered the best fit [42].

Ethical consideration

Access to the Ethiopian Mini EDHS 2019 dataset and the corresponding Global Positioning System (GPS) data was obtained through the DHS Program website (http://www.dhsprogram.com/) after registration and approval of the data request. Authorization was granted to download and use the required datasets for this analysis. Before the data collection, the materials were reviewed and approved by the National Research Ethics Review Committee in Ethiopia. The data from the 2019 EDHS did not include any personal identifiable information and written informed consent was not required.

Results

Household characteristics and overcrowding

Overcrowding was found in 73.9% of male-headed and 59.7% of female-headed households. Overcrowding was observed among 74.4% of household heads with no formal education, 70.6% primary education, 46.3% secondary education, and 53.7% with higher education. Rural households and households with five or more members accounted for 76.0% and 95.7% overcrowding respectively. Overcrowding was reported in 86.9% of single-room households and 32.1% of those with four or more rooms.

Housing and infrastructure characteristics included ownership (75.3%), rent (52.2%), or free-of-charge (57.2%), unimproved floor (71.8%), unimproved wall (72.9%), and unimproved roof (77.1%) materials were overcrowded. Overcrowding was reported in households without access to basic sanitation (70.9%) and without media exposure (74.4%). Wealth status ranged from 83.0% among the poorest to 54.4% among the richest households were overcrowded. Regionally, 72.4% of agrarians, 74.1% of pastoralists, and 54.7% of city households were overcrowded (Table 1).Table 1 Distribution of Household Characteristics by Overcrowding Status in Ethiopia in 2019.

Full size table

Prevalence of household overcrowding in Ethiopia

The pooled prevalence analysis of overcrowding across Ethiopian regions revealed significant regional disparities. The overcrowding prevalence’s were ranged from 47.3% in Addis Ababa to 83.7% in the Somali region, which indicated statistically significant differences between regions. Using a random-effects model with the Der Simonian and Laird estimator, the pooled national estimate of overcrowding was 69.62% (95% CI: 63.98–75.26). Extremely high heterogeneity among regions was reported (I² = 97.2%, p < 0.001), indicating that almost all variation was due to genuine differences in regional contexts rather than chance.Table 2 Distribution of Household Characteristics by Overcrowding Status in Ethiopia in 2019.

Full size table

Fig. 1
Fig. 1

Spatial distribution of household overcrowding

Moran’s I was used to analyze spatial autocorrelation, and the result was 0.604, indicating moderate to strong clustering of similar values across the study area. Clustering with a Z-score of 13.2 and p-value < 0.001 is statistically significant and unlikely to occur under random spatial distribution. Visualization of significance levels showed that positive Z-scores in the upper tail correspond to strongly clustered patterns, whereas negative or near-zero scores indicate dispersion or randomness(Fig. 2). This spatial pattern supports these findings, with the observed distribution closely resembling a clustered configuration.

Fig. 2
Fig. 2

Hot spot analysis of household overcrowding

The map presents a Hot Spot Analysis (Getis-Ord Gi* statistic) to explore the spatial clustering of households living in an overcrowded environment in Ethiopia. Geographical patterns were statistically significant as evidenced by marked clusters of high level of overcrowding, which was observed mainly in Addis Ababa, Dire Dawa, Harari, Gambella, Southwest Ethiopia, North SNNP and central Oromia. Cold spots (low overcrowding clusters) are found in almost all parts of Tigray, Afar, and Amhara region as well as in Northeast part of Oromia. The rest of the country (most parts of Oromia, Benishangul Gumuz, Sidama and SNNP) did not have significant spatial pattern where the observed overcrowdings were randomly distributed.

Sat scan analysis of household overcrowding

The Sat Scan analysis of household overcrowding demonstrates specific and significant cluster located at south and north central parts of the country. Oromia, Sidama, portions of SNNP and major parts of the Somali region are found to have a very statistically significant (P value = 0.005) cluster of household overcrowding(fig. 3 & table 3). In contrast, the rest of the country are not statistically significant for clustering, implying that household overcrowding is not uniformly distributed across Ethiopia.

Fig. 3
Fig. 3
Fig. 4
Fig. 4

Table 3 Significant spatial clusters of household overcrowding Ethiopia, 2019.

Full size table

Spatial interpolation of household overcrowding

The spatial distribution of household overcrowding in Ethiopia was estimated using ordinary Kriging (OK) interpolation, a geostatistical technique that produces the best linear unbiased prediction for unsampled areas. The resulting interpolation surface revealed a diverse geographical pattern of household overcrowding throughout the country. Higher predicted proportions of overcrowded households (represented in red, approximately 0.812–0.899) were concentrated in several regions, including central east and south Oromia, Eastern and Northern SNNP, Southwestern Afar, and most parts of Somalia. Areas with lower predicted overcrowding (represented in green shades, approximately 0.391–0.594) were mostly found in Addis Ababa, Dire Dawa, Western Gambella, and Northwest Amhara (Figs.5).

Fig. 5
Fig. 5

Spatial regression of the factors using ordinary least square analysis

The global model diagnostics confirmed statistical robustness with joint F-statistic (F = 5,299) and Wald statistic (χ² = 431.06, p < 0.001) indicated model significance. The Koenker statistic (χ² = 12.45, p = 0.029) indicated significant heteroscedasticity and robust standard errors were used to account for further analysis. The Jarque-Bera statistic (χ² = 21.82, p = 0.061) showed that the residuals were approximately normally distributed (Table 4). Finally, the geographically weighted regression results indicate that no educational attaining primary education, more than five family size, residing in pastoralist region, and no media exposure are significant factors of the HC.Table 4 Diagnostic statistics of ordinary least square regression results on socioeconomic and environmental factors of household overcrowding in Ethiopia (2019 EDHS).

Full size table

The global Ordinary Least Squares model accounted for 61% of the variance in the outcome indicating a good fit. The Akaike Information Criterion (AICc = −474.566) provides additional support for model parsimony. Households with no education, primary education, family size greater than five, reside in pastoralist regions and lack of media exposure were positively associated with HC. None of the variables was reported to have the Variance Inflation Factor (VIF) value beyond 7.5, the maximum allowable VIF(Table 5).Table 5 Ordinary least square regression results on socioeconomic and environmental factors of household overcrowding in Ethiopia (2019 EDHS).

Full size table

Factors associated with household overcrowding using geographically weighted regression analysis

The Geographically Weighted Regression (GWR) model had greater explanatory power than the global OLS model. After accounting for spatial heterogeneity, the model explained approximately 63% of the variation in the outcome. The AICc value (−487.23) was lower than that of the global model (−474.57), indicating that the GWR provided a better fit. The sigma value of 0.106 also suggests a good fit. Overall, the GWR model showed that relationships vary across space with improved model performance over the global OLS specification (Table 6).Table 6 Goodness-of-fit metrics of the GWR model for household overcrowding with 95% CI in Ethiopia (2019 EDHS).

Full size table

A significant west-to-east gradient in Local R² values was observed throughout Ethiopia. The model explained weakly in North to West regions including Tigray, Northwest Amhara, Benishangul-Gumuz, Gambella, and western Oromia, accounts for 41 to 49% of the variance in the data. This explanatory power gradually increases towards the eastern periphery, with relatively high values (73–80%) concentrated in the eastern Somali, Afar, central and east Oromia, Dire Dawa, and Harari regions (Fig 6). This distinct spatial pattern showed that the modeled relationships are non-stationary.

Fig. 6
Fig. 6

The geographically weighted regression revealed distinct spatial patterns for each factor. The influence of no media exposure is most potent in the Northern Tigray region, diminishing significantly in the central and Northwest highlands. No education, attaining primary education and residing in pastoralist region showed a powerful positive effect bifurcating along an east-southwest axis, strongly associating with Somali, Afar, and extending through Oromia into Gambella.

Conversely, the effect of a large family size is dominated in the north-south gradient, with the strong significant spatial pattern in the northwest parts of Ethiopia (Tigray, Amhara) and western parts of the country (Gambella and southwest Ethiopia region) (Fig 7 & 8). This demonstrates clear geographic non-stationarity in which regional contexts fundamentally shape the association between key factors and HC.

Fig. 7
Fig. 7
Fig. 8
Fig. 8

Discussion

This multi-method geospatial analysis combining hot spot analysis, Sat Scan cluster detection, Ordinary Kriging interpolation, and OLS followed by GWR provides a robust understanding of household overcrowding in Ethiopia by capturing both its spatial concentration and its region-specific factors. Sat Scan identified a highly significant non-random cluster of overcrowding in the south-central corridor, spanning Oromia, Sidama, SNNP, and the Somali region. Kriging interpolation confirmed these hotspots, while GWR revealed the diverse underlying factors, showing that the factors of overcrowding vary markedly across Ethiopia’s human and ecological landscapes41,42.

A regional based meta-analysis showed that Somali region had elevated prevalence (83.7%), followed by Oromia (78.8%) and SNNP (78.0%), consistent with cluster analyses reporting south-central and eastern regions as significant crowding hotspots. High prevalence in these areas might be due to traditional housing structures with sharing of extended families, low stock of formal housing, multiple generational co-habitation, and poor infrastructure. All of which facilitate more occupancy density and increased risk for overcrowding at the household level. Moderate level of HC with not-significant cluster analysis reports were observed in Afar (73.8%), Benishangul-Gumuz (75.0%), Amhara (67.8%), Tigray (67.7%), Harari (66.8%), Dire Dawa (61.8%), and Gambella (64.5%). All these areas have mixed settlement patterns, both urban and rural, which may be a cause of house availability and variability.

Conversely, Addis Ababa registered the lower incidence of household overcrowding at 47.3%, and this can perhaps be attributed to its urban housing stock, including relatively higher coverage of modern housing, better infrastructure, and higher proportions of apartment and condominium initiatives. In addition, increased coverage of rental housing, government housing schemes and urban planning could be contributing to lower overcrowding.

Household overcrowding remains a significant issue in Ethiopia, with a national rate of 69.6%. Higher prevalence is observed in the Somali region (78.8%), whereas lower prevalence occurs in Addis Ababa (47.3%). When compared to neighboring countries, Ethiopia’s overcrowding figures exceed Johannesburg, South Africa (57.6%)26, Djibouti’s (45%) and Eritrea’s (59%)11. Meanwhile, other sub-Saharan African countries like Niger, Angola, Kenya, Ghana, Burundi, Rwanda and Namibia report lower overcrowding rates of 48%, 43%, 41%, 38%, 38%, 30% and 20%, respectively11. This comparison indicated that Ethiopia, especially the Somali region, continues to experience significant challenges with household overcrowding.

The GWR analysis also showed that the causes of overcrowding are not consistent across the country. In the northwest of the country (Amhara, Tigray), large family size and lack of media exposure were significant factors. This could result from large family sizes that raise the number of individuals living in scanty space, thus escalating overcrowding, particularly in highland rural regions with large fertility rates. Simultaneously, poor media coverage involves structural marginalization and restricts access to information regarding housing, health, and family planning, amplifying spatial inequality in overcrowding. The influence of large family size and lack of media exposure to household overcrowding is supported by previous studies conducted in Alexandria, Egypt43 and Hong Kong44 respectively. However, this pattern did not result in statistically significant Sat Scan clusters, most likely because the effect is more evenly distributed and partially offset by housing availability.

In contrast, the central and eastern cluster including parts of Oromia, eastern Amhara, Dire Dawa, Harari, Afar and Somali appears to result from a convergence of multiple factors. Geographically weighted regression analysis identified lack of education, attaining primary education, and pastoralist livelihood as the significant spatial factors of household overcrowding. This shows how illiteracy and lack of awareness restrict access to better housing and perpetuate intergenerational poverty, as supported by previous findings43. Early-stage schooling is also found with a strong correlation, indicating that even less than minimum education learns some rudiments but fails to increase households out of structural poverty and thus maintains conditions that compel families to double up in already packed accommodation.

Similarly, pastoralist livelihoods demonstrated a clear spatial link to overcrowding, as mobile and subsistence-based economies constrain stable income and access to adequate housing infrastructure, leaving households in these areas particularly prone to congested living environments. Weak tenure systems and poor infrastructure in pastoralist areas also discourage investment in fixed housing, and cultural tradition frequently promotes extended family cohabitation, increasing person-per-room rates. Where they overlap to do so, they support one another’s effects and generate hotspots for overcrowding. This clarified that overcrowding is not only caused by demographic pressure but also caused by the joint effect of low schooling, limited economic opportunities, and housing practices associated with pastoralist livelihoods.

The consistent significance of Dire Dawa, Harari, and the Somali region in the Sat Scan cluster analysis, together with their high local R² values in GWR indicate that the measured factors account for much of the observed variation. Besides such measurable factors of overcrowding, other possible non-measurable factors like environmental displacement due to climate change, conflict, or policy might also make significant contributions. These factors at a context and structure level may sustain shortage of housing and amplify health hazards, psychosocial strain, and social disparities. Thus, crowding not only occurs because of these factors but also as factors of broader public health disparities.

Policy and public health implications

The findings reveal that reducing household overcrowding requires strategies that extend beyond demographic management and housing development. Policies must prioritize education, with special focus on expanding access to secondary and higher education in regions where lack of schooling continues to reinforce cycles of deprivation. Similarly, targeted livelihood interventions for pastoralist communities are essential to improve income stability and reduce reliance on housing practices that contribute to overcrowding.

From a public health perspective, overcrowding should be addressed as both a risk factor and health outcome. Overcrowded households increase exposure to infectious diseases, amplify psychosocial stress, and deepen social inequalities. Integrated approaches that combine housing improvements, educational opportunities, livelihood diversification, and infrastructural investments will therefore have dual benefits Special attention should be directed toward regions such as Dire Dawa, Harari, and Somali, where overlapping vulnerabilities create persistent hotspots.

Limitation of the study

Regional overcrowding prevalence was calculated using survey weights, strata, and PSUs, ensuring population-representative estimates. Cluster-level weighted proportions were used in Sat Scan, Gi*, Kriging, and GWR analyses to approximate population patterns; however, none of these spatial methods fully account for survey design, so results may be limited to areas with few clusters or highly variable weights. GWR results should be interpreted as approximations of spatially varying relationships rather than fully design-adjusted estimates. Because of DHS cluster displacement, the exact location of clusters is uncertain. Consequently, spatial patterns should be interpreted as approximate population-level trends rather than precise point estimates. Although multiple covariates were adjusted for, residual confounding may remain, as factors such as fertility frate, norms, migration, and displacement could influence the observed associations but are not fully captured in the model.

Conclusion

The magnitude of household overcrowding in Ethiopia is high, particularly in rural areas and pastoralist regions. This study also demonstrates that that lack of education, limited attainment at the primary level, no media exposure, more than five family members, living on pastoralist region significant spatial factors of household overcrowding, while spatial cluster analysis confirmed concentrated hotspots in Dire Dawa, Harari, and the Somali region. Furthermore, the consistency between cluster patterns and local model performance indicates that significant measured factors account for a substantial portion of the variation. Geographically tailored interventions including access to information in the northwest while higher education, improved livelihoods, infrastructure development, and culturally appropriate housing solutions in the east are required.

Appendix

Appendix Table A1Table A1 Description and Recoding of DHS Household Variables for Overcrowding and Associated Household Characteristics.

Full size table

Data availability

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author and/or publicly available upon request from the DHS Program website ([http://www.dhsprogram.com/](http:/www.dhsprogram.com/?utm_source=chatgpt.com))

References

  1. Grabovyi, P. G. et al. Housing quality and health; an evaluation of influencing variables. E3S Web of Conferences 535, (2024). https://doi.org/10.1051/e3sconf/202453503004
  2. Holden, K. A., Lee, A. R., Hawcutt, D. B. & Sinha, I. P. The impact of poor housing and indoor air quality on respiratory health in children. Breathe (Sheff). 19, 230058. https://doi.org/10.1183/20734735.0058-2023 (2023).Article PubMed PubMed Central Google Scholar 
  3. Adler, N. E. & Newman, K. Socioeconomic disparities in health: pathways and policies. Health Aff (Millwood). 21, 60–76. https://doi.org/10.1377/hlthaff.21.2.60 (2002).Article PubMed Google Scholar 
  4. Krieger, J. & Higgins, D. L. Housing and health: time again for public health action. Am. J. Public. Health. 92, 758–768. https://doi.org/10.2105/ajph.92.5.758 (2002).Article PubMed PubMed Central Google Scholar 
  5. Rana, K. Sustainable housing as a social determinant of health and wellbeing. Sustainability 17, 7519 (2025).Article ADS CAS Google Scholar 
  6. Swope, C. B. & Hernandez, D. Housing as a determinant of health equity: A conceptual model. Soc. Sci. Med. 243, 112571. https://doi.org/10.1016/j.socscimed.2019.112571 (2019).Article PubMed PubMed Central Google Scholar 
  7. Rolfe, S. et al. Housing as a social determinant of health and wellbeing: developing an empirically-informed realist theoretical framework. BMC Public. Health. 20, 1138. https://doi.org/10.1186/s12889-020-09224-0 (2020).Article PubMed PubMed Central Google Scholar 
  8. WHO. World Health Organization (2018).
  9. Baker, M., Das, D., Venugopal, K. & Howden-Chapman, P. Tuberculosis associated with household crowding in a developed country. J. Epidemiol. Community Health. 62, 715–721. https://doi.org/10.1136/jech.2007.063610 (2008).Article CAS PubMed Google Scholar 
  10. Lee, J. Y., Kwon, N., Goo, G. Y. & Cho, S. I. Inadequate housing and pulmonary tuberculosis: a systematic review. BMC Public. Health. 22, 622. https://doi.org/10.1186/s12889-022-12879-6 (2022).Article PubMed PubMed Central Google Scholar 
  11. Chipeta, M. G. et al. Mapping local variation in household overcrowding across Africa from 2000 to 2018: A modelling study. The Lancet Planetary Health 6, e670–e681. https://doi.org/10.1016/S2542-5196(22)00149-8 (2022).Article PubMed PubMed Central Google Scholar 
  12. Islam, M., Sultana, Z. Z., Iqbal, A., Ali, M. & Hossain, A. Effect of in-house crowding on childhood hospital admissions for acute respiratory infection: A matched case–control study in Bangladesh. Int. J. Infect. Dis. 105, 639–645. https://doi.org/10.1016/j.ijid.2021.03.002 (2021).Article PubMed Google Scholar 
  13. Howie, S. R. et al. Childhood pneumonia and crowding, bed-sharing and nutrition: A case-control study from The Gambia. Int J Tuberc Lung Dis 20, 1405–1415. https://doi.org/10.5588/ijtld.15.0993 (2016).Article CAS PubMed PubMed Central Google Scholar 
  14. Baker, M. et al. Household crowding a major risk factor for epidemic meningococcal disease in Auckland children. Pediatr Infect Dis J 19, 983–990. https://doi.org/10.1097/00006454-200010000-00009 (2000).Article CAS PubMed Google Scholar 
  15. Deutch, S. et al. Crowding as a risk factor of meningococcal disease in Danish preschool children: A nationwide population-based case-control study. Scand J Infect Dis 36, 20–23. https://doi.org/10.1080/00365540310017500 (2004).Article PubMed Google Scholar 
  16. Baker, M. G. et al. Risk factors for acute rheumatic fever: A case-control study. The Lancet Regional Health – Western Pacific 26, 100508. https://doi.org/10.1016/j.lanwpc.2022.100508 (2022).Article PubMed PubMed Central Google Scholar 
  17. Jaine, R., Baker, M. & Venugopal, K. Acute rheumatic fever associated with household crowding in a developed country. Pediatr. Infect. Dis. J. 30, 315–319. https://doi.org/10.1097/INF.0b013e3181fbd85b (2011).Article PubMed Google Scholar 
  18. Hosseini, L. J. et al. Household overcrowding in Iran, a low-middle-income country: How major of a public health concern is it?. J Prev Med Public Health 54, 73–80. https://doi.org/10.3961/jpmph.20.568 (2021).Article PubMed PubMed Central Google Scholar 
  19. McNicholas, A., Lennon, D., Crampton, P. & Howden-Chapman, P. Overcrowding and infectious diseases–When will we learn the lessons of our past?. N. Z. Med. J. 113, 453–454 (2000).CAS PubMed Google Scholar 
  20. McIver, L. et al. Overcrowding and Non–Communicable Diseases: the Triple Whammy of Tuberculosis Transmission Risk in Pacific Atoll Countries. Annals ACTM: Int. J. Trop. Travel Med. 16, 57–61 (2015).Google Scholar 
  21. Goux, D. & Maurin, E. The effect of overcrowded housing on children’s performance at school. J. Public Econ. 89, 797–819. https://doi.org/10.1016/j.jpubeco.2004.06.005 (2005).Article Google Scholar 
  22. Solari, C. D. & Mare, R. D. Housing crowding effects on children’s wellbeing. Soc. Sci. Res. 41, 464–476. https://doi.org/10.1016/j.ssresearch.2011.09.012 (2012).Article PubMed Google Scholar 
  23. Marsh, R. et al. The association between crowding within households and behavioural problems in children: Longitudinal data from the Southampton Women’s Survey. Paediatr. Perinat. Epidemiol. 33, 195–203. https://doi.org/10.1111/ppe.12550 (2019).Article PubMed PubMed Central Google Scholar 
  24. Ruiz-Tagle, J. & Urria, I. Household overcrowding trajectories and mental well-being. Soc. Sci. Med. 296, 114051. https://doi.org/10.1016/j.socscimed.2021.114051 (2022).Article PubMed Google Scholar 
  25. USHUD. (2007).
  26. Nkosi, V., Haman, T., Naicker, N. & Mathee, A. Overcrowding and health in two impoverished suburbs of Johannesburg, South Africa. BMC Public Health 19, 1358. https://doi.org/10.1186/s12889-019-7665-5 (2019).Article PubMed PubMed Central Google Scholar 
  27. Gu, D., Andreev, K. & Dupre, M. E. Major trends in population growth around the world. China CDC Wkly. 3, 604–613. https://doi.org/10.46234/ccdcw2021.160 (2021).Article PubMed PubMed Central Google Scholar 
  28. Vollset, S. E. et al. Fertility, mortality, migration, and population scenarios for 195 countries and territories from 2017 to 2100: A forecasting analysis for the Global Burden of Disease Study. Lancet 396, 1285–1306. https://doi.org/10.1016/S0140-6736(20)30677-2 (2020).Article PubMed PubMed Central Google Scholar 
  29. Zhang, Y. & Urbanization Inequality, and Poverty in the People’s Republic of China. (2016).
  30. Yetunderonke, O. A. Study of Housing Adequacy of Multi-Habited Houses in a Typical Nigerian Town. Int. J. Adv. Multidisciplinary Social Sci. 1 https://doi.org/10.5923/j.jamss.20150101.02 (2015).
  31. Fiadzo, E. On the Estimation of Determinants of Housing Quality (The Case of Ghana, 2011).
  32. Yallew, W. W. et al. Household sanitation and crowding status in Addis Health and Demographic Surveillance System (Addis-HDSS) in Addis Ababa, Ethiopia. Ethiop. J. Health Sci. 34, 84–90. https://doi.org/10.4314/ejhs.v34i2.3S (2024).Article PubMed PubMed Central Google Scholar 
  33. Larsen, L., Yeshitela, K., Mulatu, T., Seifu, S. & Desta, H. The impact of rapid urbanization and public housing development on urban form and density in Addis Ababa, Ethiopia. Land 8, 66. https://doi.org/10.3390/land8040066 (2019).Article Google Scholar 
  34. ICF-International. Mini Demographic and Health Survey 2019 Report. (2021).
  35. Dearden, K. A. et al. Children with access to improved sanitation but not improved water are at lower risk of stunting compared to children without access: A cohort study in Ethiopia, India, Peru, and Vietnam. BMC Public Health 17, 1–19 (2017).Article Google Scholar 
  36. ICF, E. P. & Rockville H. I. E. a. Ethiopia Mini Demographic and Health Survey 2019: Final Report., (Maryland, USA: EPHI and ICF., (2021).
  37. WHO/UNICEF. Progress on household drinking water, sanitation and hygiene 2000–2020: five years into the SDGs. Report No. 9240030840 (World Health Organization, 2021).
  38. Colston, J. M. et al. Spatial variation in housing construction material in low- and middle-income countries: A Bayesian spatial prediction model of a key infectious diseases risk factor and social determinant of health. PLOS Glob. Public Health 4, e0003338. https://doi.org/10.1371/journal.pgph.0003338 (2024).Article PubMed PubMed Central Google Scholar 
  39. Florey, L. & Taylor, C. Using household survey data to explore the effects of improved housing conditions on malaria infection in children in Sub-Saharan Africa (ICF International, 2016).Google Scholar 
  40. Atangana, E. & Oberholster, P. J. Assessment of water, sanitation, and hygiene target and theoretical modeling to determine sanitation success in sub-Saharan Africa. Environ. Dev. Sustain. 25, 13353–13377 (2023).Article Google Scholar 
  41. Bai, H. & Li, S. Geo-spatial Data Analysis, Quality Assessment and Visualization. (2008).
  42. Zhang, K. & Chen, M. Multi-method analysis of urban green space accessibility: Influences of land use, greenery types, and individual characteristics factors. Urban For. Urban Green. 96, 128366. https://doi.org/10.1016/j.ufug.2024.128366 (2024).Article MathSciNet Google Scholar 
  43. Mansour, S. Spatial modeling of residential crowding in Alexandria Governorate, Egypt: A geographically weighted regression (GWR) technique. J. Geographic Inform. Syst. 07, 369–383. https://doi.org/10.4236/jgis.2015.74029 (2015).Article ADS Google Scholar 
  44. Kung, Y. The influence of media exposure on the willingness of homeowners to participate in property management (The University of Hong Kong, 2024).

Download references

Acknowledgements

The authors gratefully acknowledge the Demographic and Health Surveys (DHS) program for granting access to the data used in this analysis.

Funding

The author(s) declared that no financial support was received for the research, authorship, and/or publication of this article.

Author information

Authors and Affiliations

  1. Department of Environmental Health, College of Medicine and Health Science, Wollo University, Dessie, EthiopiaAwoke Keleb
  2. Department of Biomedical Science, College of Medicine and Health Science, Wollo University, Dessie, EthiopiaAltaseb Beyene Kassaw & Gosa Mankelkl
  3. Department of Occupational Health and Safety, College of Medicine and Health Science, Wollo University, 1145, Dessie, EthiopiaAnmut Endalkachew Bezie
  4. School of Public Health, College of Medicine and Health Science, Wollo University, Dessie, EthiopiaAdisu Meles Kabtyimer
  5. Department of Environmental Engineering, School of Mechanical and Chemical Engineering, Kombolcha Institute of Technology, Wollo University, Kombolcha, EthiopiaWubalem Amare
  6. Department of Health Informatics, College of Medicine and Health Science, Wollo University, Dessie, EthiopiaHalid Worku Jemil

Contributions

AK and AEB were responsible for data curation and formal analysis, while AK led the investigation. The methodology was developed collaboratively by AK, ABK, AEB, GM, AMK, WA, and HWJ. Software implementation was handled by AK. The original draft of the manuscript was prepared by AK, ABK, WA and AEB. With all authors contributing to subsequent review and editing.

Corresponding author

Correspondence to Awoke Keleb.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Keleb, A., Kassaw, A.B., Bezie, A.E. et al. Geospatial patterns and socio-environmental factors of household overcrowding in Ethiopia: Evidence from 2019 Ethiopian demographic and health survey data. Sci Rep 16, 8504 (2026). https://doi.org/10.1038/s41598-026-42516-w

Download citation

Share this article