20 January, 2023 
In an attempt to understand open defecation in four focus states in the aftermath of the Swachh Bharat Mission, Vyas and Gupta evaluate the findings of the NFHS-5. They note that the NFHS is likely to underestimate open defection rates due to their use of data collected at the household level, and the possibility of response bias. After adjusting the estimates, they find that about half of rural Indians in the focus states defecated in the open in 2019-21.
After five years of the Swachh Bharat Mission (SBM), India was declared open defecation free in October 2019. Because of the constraints posed by Covid-19 and the associated lockdown, no nationally representative survey data were made available to assess this claim until recently. We use recently released micro-data from India’s National Family Health Survey 5 (NFHS-5), which surveyed households between 2019 and 2021, to understand open defecation in rural India since SBM.
We are particularly interested in understanding open defecation in rural Bihar, Madhya Pradesh, Rajasthan, and Uttar Pradesh, which we collectively call the focus states. These states represent about two-fifths of India’s rural population, and are home to most of the rural households that were defecating in the open before the SBM. Since 2014, along with coauthors, we have done both quantitative data collection, as well as qualitative field work in these states, focusing on sanitation attitudes, preferences, and behaviors (Coffey et al. 2014, Coffey et al. 2017, Gupta et al. 2019, Gupta et al. 2020).
Between 2015-16 and 2019-21, the NFHS estimated a reduction in open defecation of 28 percentage points among all rural households, and 36 percentage points among rural households in the focus states. In 2019-21, about one-quarter of all rural households, and about one-third of rural households in the focus states, continued to defecate in the open.
This gives one answer to the question posed in the title of this article. In the rest of this article, we discuss why this might underestimate open defecation in rural India. We use the NFHS data to make more sophisticated estimates of rural open defecation in the focus states, and using different methods, we estimate that about half of rural Indians in the focus states defecated in the open between 2019 and 2021.
Why the NFHS is likely to underestimate open defecation
One reason why the NFHS is likely to underestimate open defecation is because it asks a household-level question about individual behaviour. For each household, the survey asks: “What kind of toilet facility do members of your household usually use?” Surveyors classify responses into different types of toilets, or open defecation. Prior research in rural India has found that even among latrine-owning households, it is common for some individuals to use the latrine and for others to defecate in the open (Coffey et al. 2014). Using a randomised survey experiment, Vyas et al. (2019) find that the household-level question asked in the NFHS finds 20 percentage points less open defecation than an individual-level question.
Another reason why the NFHS-5 in particular might underestimate open defecation is because it followed an intensive, high-profile, and coercive campaign to eliminate open defecation (Gupta et al. 2019). This might have caused respondents to provide answers that they perceived to be more socially desirable or partly correct, even if inaccurate on the whole.
The NFHS-5 also asked some additional questions for those who responded “open defecation” to the original question. Surveyors were instructed to ask: “Do members of your household have access to a toilet facility?” and “What kind of toilet facility do members of your household have access to?” These additional questions were newly added to the NFHS-5 questionnaire, and do not appear in prior NFHS questionnaires. Although we were not involved in designing the NFHS-5 questionnaire, it is likely that these follow-up questions were included to capture latrine construction from the SBM. In practice, these follow-up questions might have discouraged a respondent from responding “open defecation” to the original question, or might have encouraged the respondent to change her response.
We cannot observe these types of response bias directly in the NFHS. However, a comparison between the NFHS-5 and the Survey of Rural Sanitation and Solid Fuel Use (RSFU), a household survey representative of the focus states which we carried out in 2018, suggests that the NFHS-5 might have underestimated open defecation. We make comparisons over two different indicators.
First, as shown in Table 1, the NFHS finds much less open defecation among latrine owners compared to the RSFU.1 According to the RSFU, about one-fifth of individuals in latrine-owning households defecated in the open in 2018, a statistic which has remained very stable since a 2014 survey of the same households. According to NFHS-5, however, only about 2% of latrine-owning households defecated in the open.
Table 1. Open defecation among latrine owners
| Sample | Open defecation among: | Latrine access among households | ||
|---|---|---|---|---|
| Individuals in latrine-owning households | Households with latrine access |
|
||
|
Survey |
RSFU (2018) |
NFHS-5 (2019-21) |
RSFU (2018) |
NFHS-5 (2019-21) |
|
|
|
|
||
|
Rural India |
-- |
3% |
-- |
75% |
|
|
|
|
||
|
Rural focus states |
23% |
2% |
71% |
67% |
|
|
|
|
|
|
|
Rajasthan |
40% |
2% |
78% |
72% |
|
Uttar Pradesh |
21% |
2% |
73% |
73% |
|
Bihar |
21% |
1% |
50% |
55% |
|
Madhya Pradesh |
16% |
4% |
90% |
69% |
Source: RSFU estimates from Gupta et al. (2020), and NFHS-5 (2019-21).
Is it possible that open defecation among latrine owners could have reduced so dramatically between 2018 and 2019-21? By the time the RSFU started data collection, both Madhya Pradesh and Rajasthan had already declared their states open defecation free and work on the SBM had largely come to an end. Yet, 16% of people in latrine-owning households in Madhya Pradesh, and 40% in Rajasthan, continued to defecate in the open. This suggests that the NFHS-5 likely underestimated open defecation among latrine owners.
Second, an examination of the types of latrines recorded in response to the original NFHS question shows variation across states that is at odds with the RSFU. Both surveys coded single-pit latrines, twin-pit latrines, and septic tanks separately, along with other variations on toilet construction. The RSFU found that about one-quarter of the latrines in the focus states were twin-pit latrines. In the NFHS-5 data, though, Uttar Pradesh is the only one of the focus states that recorded a substantial fraction of twin-pit latrines. The other states show much lower proportions of twin-pit latrines. Relative to the NFHS, the 2018 RSFU found higher proportions of twin-pit latrines in Bihar, Madhya Pradesh, and Rajasthan.
In our field work in 2018, we found that local officials in Uttar Pradesh had placed a greater emphasis on educating people about twin-pit latrines relative to the other states. This ensured that the latrines that were constructed were more likely to be twin-pit latrines, and might explain some of the difference across states in the types of latrines observed in the subsequent NFHS. The focus on twin-pits by the Uttar Pradesh state government might have influenced surveyor training, contributed to greater social desirability bias, and/or led to greater awareness among respondents, such that twin-pit latrines were more likely to be observed in the state.
Incidence of rural open defecation in the focus states
We estimate rural open defecation in the focus states in several ways. The first estimate requires no computation: it is simply the fraction of households in the NFHS-5 that responded “open defecation” in the survey. This is shown in column 2 of Table 2. For the reasons we described above, though, this is likely an underestimate of open defecation.
In columns 3 and 4 of Table 2, we compute estimates using two different adjustment methods. The first method adjusts the NFHS-5 estimates based on the findings from Vyas et al. (2019). For that study, we worked with a team of researchers to organise a randomised survey experiment in rural Bihar, Odisha, Karnataka, and Gujarat. Half the households received the NFHS question, and half received an individual-level question about open defecation. We found that the NFHS question underestimated open defecation by about 20 percentage points on average, and by 10 percentage points in the state that showed the smallest difference between the two questions. Column 3 shows a range for rural open defecation based on these findings, in which we adjust the NFHS-5 estimates by 10-20 percentage points.2 Using this method, about half of rural households in the focus states defecated in the open in 2019-21.
Table 2. Adjusted estimates of open defecation in rural north India
| Survey | RSFU (2018) | NFHS-5 (2019-21) | ||||
|---|---|---|---|---|---|---|
| Sample unit | Individual | Household | Individual | Individual | ||
| Computation method | None | None | Adjustment based on Vyas et al. (2019) | Prediction based on RSFU model estimates | ||
|
|
(1) |
(2) |
(3) |
(4) |
||
|
|
|
|
||||
|
Rural India |
-- |
27% |
37-47% |
-- |
||
|
|
|
|
||||
|
Rural focus states |
44% |
34% |
44-54% |
45% |
||
|
|
|
|
|
|
||
|
Rajasthan |
53% |
29% |
-- |
51% |
||
|
UP |
39% |
29% |
-- |
38% |
||
|
Bihar |
60% |
45% |
-- |
55% |
||
|
MP |
25% |
34% |
-- |
38% |
||
Source: Author calculations based on NFHS-5 (2019-21) and RSFU (2018).
Note: We do not show estimates in column 3 for states separately because of the imprecision of the estimates.
The second method we use predicts open defecation among individuals over the age of two in NFHS-5 households based on a simple logistic model where open defecation in the RSFU is a function of household and individual characteristics. The characteristics we use in the model include: age, household size, the type of latrine the household owns (pit latrine, septic tank, other, or none), and state, each interacted with an indicator for female, as well as ownership of each of seven assets.3 If we code an individual’s open defecation prediction based on a predicted probability4 greater than 0.50, then this model accurately predicts the behavior of 82% of individuals over age two in the RSFU. Applying this model to individuals in the NFHS gives a prediction for rural open defecation in the focus states of a little less than one half.
Both adjustment methods yield similar results: about half of rural Indians in the focus states defecated in the open in 2019-21. This is very similar to the 2018 RSFU estimate. In Coffey et al. (2021), our colleagues wrote about what could be learned about rural open defecation in the focus states from the NFHS District Level Fact Sheets from Phase 1, which provided estimates of improved sanitation for Bihar, and came up with a similar estimate of rural open defecation in the focus states: about half.
Conclusion
Relative to the 70% of rural households that were estimated to defecate in the open in the focus states in 2015-16, all estimates of open defecation in 2019-21 represent a considerable decline. Depending on the precise estimate used for 2019-21, rural open defecation in the focus states declined by 6-9 percentage points per year. This is a rapid acceleration compared to the 2 percentage point annual rate of decline between 2005-06 and 2015-16 (Coffey and Spears 2018). However, open defecation is far from having been eliminated in rural India, and the work to encourage latrine use must continue.
Notes:
- To estimate open defecation among latrine owners in the NFHS-5, we use the newly added questions. Households are classified as defecating in the open if they chose “open defecation” in response to the original question, but we classify a household as having access to a latrine if the respondent either said they use a latrine in response to the original question, or said they have access to a latrine in response to the additional questions. In the RSFU, we separately measured latrine ownership and individual-level latrine use.
- This adjustment can best be thought of as measuring household open defecation as a variable that can take on fraction values in the range 0-1, rather than a binary variable.
- The seven assets include: watch, electricity, bicycle, mobile phone, television, motorcycle, and refrigerator.
- In a logistic regression model, the predicted probability is the probability that the individual defecates in the open.
Further Reading
- Coffey, Diane, Aashish Gupta, Payal Hathi, Nidhi Khurana, Nikhil Srivastav, Sangita Vyas and Dean Spears (2014), “Open defecation: evidence from a new survey in rural north India”, Economic and Political Weekly, 49(38): 43-45.
- Coffey, Diane, Aashish Gupta, Payal Hathi, Dean Spears, Nikhil Srivastav and Sangita Vyas (2014), “Understanding open defecation in rural India: Untouchability, Pollution, and Latrine Pits”, Economic and Political Weekly, 52(1): 59-66.
- Coffey, Diane and Dean Spears (2018), “Open defecation in rural India, 2015-2016: Levels and trends in the NFHS-4”, Economic and Political Weekly, 53(9): 10-13.
- Coffey, D, N Franz and D Spears (2021), ‘What can we learn about Swachh Bharat Mission from NFHS-5 factsheets?’, Ideas for India, 2 February.
- Gupta, A, N Khalid, P Hathi, N Srivastav, S Vyas and D Coffey (2019), ‘Coercion, construction, and ODF ‘paper pe’’, The India Forum, 5 March
- Gupta, Aashish, Nazar Khalid, Devashish Deshpande, Payal Hathi, Avani Kapur, Nikhil Srivastav, Sangita Vyas, Dean Spears and Diane Coffey (2020), “Revisiting Open Defecation: Evidence from a Panel Survey of Rural North India”, Economic and Political Weekly, 55(21): 55.
- Vyas, Sangita, Nikhil Srivastav, Divya Mary, Neeta Goel, Sujatha Srinivasan, Ajaykumar Tannirkulam, Radu Ban, Dean Spears and Diane Coffey (2019), “Measuring open defecation in India using survey questions: Evidence from a randomized survey experiment”, BMJ Open.
Comments will be held for moderation. Your contact information will not be made public.