1 Introduction

Metsimotlhabe is a village in Kweneng District, about 20 km north-west of Gaborone on the road to Molepolole. Like other villages around the capital, it has grown rapidly as people look for cheaper plots within commuting distance of the city. Many of the new plots are not yet connected to the water network, so in the same village some households have a tap in the yard while their neighbours carry water in 20-litre containers from a public standpipe. This makes it a useful place to study a classic geographical relationship: how the friction of distance affects the way people use a resource.

The amount of water a household uses, and how far it is carried, matter for health: across 26 sub-Saharan African countries, a 15-minute shorter one-way walk to water was associated with a 41 % average relative reduction in diarrhoea among children under five (Pickering and Davis, 2012). Howard and Bartram (2003) describe "basic access" as collection from 100 to 1,000 m away, or 5 to 30 minutes in total, and state that average use at this level is "unlikely to exceed 20 l/c/d" (litres per capita per day). With one tap on the plot ("intermediate access") average use rises to about 50 litres, and the second edition of the report keeps essentially the same service levels (Howard et al., 2020). The WHO/UNICEF Joint Monitoring Programme (JMP) uses time rather than distance: a source counts as "basic" if collection takes no more than 30 minutes for a round trip, including queuing (JMP, n.d.). Both frameworks assume that the further people must walk, the less water they use. Yet early research in East Africa found that if water has to be carried, "the quantity brought home varies little for sources between 30 metres and 1000 metres" (Thompson et al., 2001, summarising White, Bradley and White, 1972). Whether distance really controls water use within a village is therefore an open question.

The question also matters beyond one village. Target 6.1 of the Sustainable Development Goals is to "achieve universal and equitable access to safe and affordable drinking water for all" by 2030 (United Nations, 2015). Botswana is semi-arid, and in peri-urban villages the main choice facing planners is whether to build more public standpipes, which shortens the walk, or to connect more plots, which removes the walk altogether. If distance is the main control on water use, more standpipes would be the cheaper answer; if the type of access matters more, connections would do more good. In geographical terms, I am testing whether water use follows a simple distance-decay pattern, in which the "friction of distance" steadily reduces interaction with a resource, or whether there is a sharp break between households that carry water and those that do not.

This essay investigates the question: To what extent does distance to the nearest water point explain daily water use per person in Metsimotlhabe, near Gaborone? The scope is one part of the village, 48 households along four transects, surveyed in April 2026. I measured distance with GPS and estimated water use with a questionnaire, water meter readings and seven-day water diaries. I then tested the relationship with Spearman's rank correlation and compared households with and without yard taps. Water quality, sanitation and commercial use are outside the scope.

This essay will argue that distance has a real and statistically significant effect on water use among households that collect from standpipes, but that the type of access, a yard tap or a standpipe, is a far stronger explanation. Distance therefore explains water use only to a limited extent, and mainly within one group of households.

2 Methodology

My first plan was to survey households near my aunt's plot, but my supervisor pointed out that this would over-represent one street and one income group. Instead I used systematic sampling along four transects that run from the main road towards the newer plots on the edge of the village (Figure 1). Along each transect I approached every third household until I had 12 completed questionnaires; when nobody was at home I moved to the next plot. This produced 48 households: 36 that collect water from a public standpipe and 12 with a yard tap. The transects cross both the older core and the newer edge, so they sample a wide range of distances.

For each standpipe household I measured the distance to the nearest standpipe as a straight line between two GPS points, one at the gate and one at the standpipe, using a phone app. Yard-tap households were given a distance of 0 m.

Water use was measured in three ways. In the questionnaire (Appendix A), standpipe households reported how many 20-litre containers they usually collect each day; I multiplied this by 20 and divided by household size. For yard-tap households I read the water meter at the start and end of the same seven-day period, with the owner's permission, and divided by household size and by seven. Finally, 20 households (15 standpipe and 5 yard-tap) kept a seven-day water diary, ticking each container or bucket used, so that I could check the other two estimates. I also asked about monthly spending on water, interruptions to supply in the past month and, for standpipe households, why they were not connected.

The village development committee gave permission for the survey, every respondent gave verbal consent after I explained the purpose, and no names or plot numbers were recorded; households are identified only by codes (H01 to H48).

To test the relationship between distance and water use I chose Spearman's rank correlation coefficient (r_s). It does not require the data to be normally distributed or the relationship to be linear, which suits a small sample with estimated values. To compare yard-tap and standpipe households I used the Mann–Whitney U test for the same reason. Both tests were calculated in a spreadsheet and checked by hand for the Spearman value.

My null hypothesis (H₀) was that there is no relationship between distance to the nearest standpipe and water use per person among standpipe households; my alternative hypothesis (H₁) was that water use falls as distance increases. Before using the container counts I checked the containers themselves in ten households. Most were the standard 20-litre plastic containers, but two households used 25-litre containers, so for those I multiplied by 25 instead of 20. This is a small correction, but without it two households would have been recorded as using less water than they really did.

3 Distance and water use among standpipe users

The first part of the argument is that distance does affect water use, at least among households that carry their water. Table 1 groups the 36 standpipe households into three distance bands and compares them with the yard-tap households.

Among standpipe users, mean use falls in every band as distance increases, from 27.0 litres within 200 m to 20.4 litres beyond 400 m. The ranges overlap, so the bands are not completely separate, but the direction is consistent. The 31 households more than 200 m from a standpipe averaged 22.3 litres, just above the 20 litres that Howard and Bartram (2003) associate with basic access, and 11 of the 36 standpipe households used less than 20 litres per person per day. Only one of those 11 lives within 400 m of a standpipe.

Figure 2 shows the same pattern for individual households. To test it, I ranked the 36 standpipe households by distance and by water use and calculated Spearman's rank correlation coefficient:

The result, r_s = −0.71, shows a strong negative relationship. For n = 36 the critical value at the 0.01 significance level (two-tailed) is about 0.42, so I can reject the null hypothesis that there is no relationship between distance and water use. (Distances were rounded to the nearest 5 m, so 13 households share a distance with at least one other; I gave tied values the average of their ranks, and the spreadsheet's tie-corrected coefficient, −0.707, rounds to the same value.) The diary households confirm the result: for the 15 standpipe households that kept a diary, r_s = −0.83.

I also checked whether the pattern holds on each transect separately, since a single unusual area could produce a misleading overall result. The coefficients for the four transects are all negative: −0.37 (T1), −0.85 (T2), −0.63 (T3) and −0.75 (T4). With only nine standpipe households on each transect, only T2 and T4 are significant at the 0.05 level (two-tailed), so these results support the overall pattern rather than prove it independently. The weakest result, on T1, is probably because most of its households are between 445 m and 635 m from the standpipe, so there is little range in distance for the correlation to detect.

However, a strong correlation is not the same as a large effect. The best-fit line in Figure 2 has a gradient of about −1.75 litres per person per day for every 100 m. Across the full range of my data, from 75 m to 635 m, it predicts a fall from about 28.8 to 19.0 litres, roughly 10 litres. Distance therefore ranks the standpipe households very consistently, but the amount it changes is modest.

This result partly contradicts the "plateau" of consumption described by Howard and Bartram (2003), in which there is "little change in quantity of water collected" between about 100 m and 1,000 m (Cairncross and Feachem, 1993, cited in Howard and Bartram, 2003). My households lie almost entirely inside that range, yet use still fell by about a third. One possible explanation is that in Metsimotlhabe a household 600 m from a standpipe can see neighbours with yard taps and may buy or borrow water from them on some days, which my questionnaire did not record. Another is that the water is carried mostly by women and children (Section 4), for whom each extra 100 m with a 20 kg container is a real burden. The contradiction is therefore real, but its size may depend on local conditions.

Squaring r_s gives 0.50, so distance explains about half of the variation in water use among standpipe users. The other half must come from factors such as household size and the cost of water, which are examined in Section 5. So far, then, the evidence supports the first half of my argument: among households that carry water, distance matters, and it matters consistently, but it changes water use by litres rather than by tens of litres.

4 Yard taps and standpipes: the type of access

The second part of the argument is that the type of access explains much more of the difference than distance. The 12 yard-tap households used a mean of 61.9 litres per person per day, 2.7 times the standpipe mean of 23.0 litres and almost three times the 22.3 litres of households more than 200 m away. Every yard-tap household used more water than every standpipe household: the lowest tap value (46.8 litres) is well above the highest standpipe value (32.1 litres). The Mann–Whitney U test therefore gives U = 0, the most extreme result possible, and the difference is significant at the 0.001 level.

Figure 3 shows the means for the four groups. The yard-tap bar is 61.9 litres, the 75–200 m bar is 27.0 litres, the 201–400 m bar is 25.0 litres and the 401–635 m bar is 20.4 litres.

One way to judge "to what extent" is to compare how much of the variation in water use across all 48 households each factor accounts for. A linear model using distance alone (with taps at 0 m) gives R² = 0.68, because yard-tap households are both at 0 m and high users. A model using only the type of access gives R² = 0.90. Adding distance to type of access raises R² only to 0.92. In other words, once we know whether a household has a tap, knowing its distance from a standpipe adds very little. Most of the apparent effect of "distance" across the whole village is really the effect of having water on the plot.

This step change matches the pattern in East Africa. In the Drawers of Water II survey of 1997–99, piped households used 66 litres per person per day and unpiped households 19.7 litres, so piped households used more than three times as much (Thompson et al., 2001). My figures are remarkably similar, even though Metsimotlhabe is in a different country and the survey is more than 25 years later. It also fits Howard and Bartram's (2003) service levels, in which water on the plot raises average use from about 20 to about 50 litres. My yard-tap households used more than 50 litres, which may be because some of them water gardens, a point I evaluate in Section 6.

The questionnaire also suggests why the step is so large. In 22 of the 36 standpipe households the main collector was an adult woman, in 8 it was a child or teenager, in 4 an adult man and in 2 the task was shared. This matches survey data from 24 sub-Saharan African countries, where adult women were the main collectors in every country among households needing more than 30 minutes to collect water (Graham, Hirai and Kim, 2016). Each 20-litre container weighs about 20 kg, so the effort of each trip falls mainly on women and children. Standpipe households described saving water by reusing washing water and bathing from a basin, while several yard-tap households mentioned washing machines, flush toilets and gardens. Once water is on the plot, new uses become possible, and these uses, rather than the absence of a short walk, explain most of the extra 39 litres. This is why a tap produces a step change, not just a continuation of the distance gradient down to 0 m.

A possible objection is that yard-tap households are simply wealthier and would use more water wherever they lived. My questionnaire did not ask about income, so I cannot rule this out. However, the large and consistent step between the two groups, and the fact that it appears in very different places, suggests that access itself is an important cause: when water does not have to be carried, the limit on use moves from the effort of carrying to the cost of the bill. The second half of my argument is therefore supported: the type of access, not distance, is the main explanation of the difference in water use across the village.

5 Other factors

If distance explains only part of the variation among standpipe users, other factors must explain the rest. I tested three that respondents mentioned: household size, the cost of water and interruptions to supply.

A counter-argument is that larger households share water more efficiently, for example by cooking one meal for everyone, so that per-person use falls as households grow. My data partly support this. Among standpipe users the correlation between household size and water use per person is weak and negative (r_s = −0.33), which is just short of significance at the 0.05 level for n = 36. Households of six or more used about 3 litres less per person than smaller ones. However, Table 2 shows that the gap between yard-tap and standpipe households is between 36 and 39 litres in every size group. Household size cannot explain the difference between the two types of access, and since it is not related to distance (r_s = −0.08), it does not distort the distance result either.

The cost of water could also matter. Twelve of the 36 standpipe households paid nothing for water in the previous month, and the other 24 paid on average P21, mostly for prepaid standpipe tokens. There was no relationship between spending and water use among standpipe users (r_s = 0.05). The cost that seems to matter is not the price of water but the price of a connection: 21 of the 36 standpipe households gave cost as the main reason why they had no yard tap, 8 said their plot was not yet serviced by the network and 4 were renting. This shifts the question from "how far do people walk?" to "why do some households not have a tap?"

Connection costs have been a concern in Botswana since the Water Utilities Corporation (WUC) took over village water supply. In 2013 residents complained about a P1,500 connection fee, and the WUC was installing prepaid standpipes in villages (Keakabetse, 2013). This newspaper report is more than ten years old and the fee may have changed, so I use it only to show that the barrier is long-standing, not to give a current price.

A final counter-argument is that supply interruptions affect everyone, whatever the distance, so they could mask the effect of distance. Interruptions in the previous month were reported by 31 of the 48 households: 23 of the 36 standpipe households and 8 of the 12 yard-tap households. Among standpipe users, those who reported interruptions used 23.2 litres per person per day and those who did not used 22.5 litres, almost the same. Yard-tap households with interruptions used less (60.4 against 64.8 litres), probably because a tap that runs dry cannot be used, while standpipe users already store water in containers. Interruptions reduce use slightly for tap users, but they do not explain either the distance gradient or the gap between the two groups.

6 Discussion and evaluation of the method

The conclusions depend on how reliable my estimates of water use are. Table 3 compares the questionnaire and meter estimates with the seven-day diaries.

For standpipe households the diaries were about 5 % lower than the questionnaire answers, which suggests that people slightly overestimate the number of containers they "usually" collect. The difference is small and similar at all distances, and the ranking of households hardly changes (r_s = 0.98 between the two estimates for all 20 diary households), so it does not affect the correlation. A weakness is that households knew they were being studied and may have been more careful while keeping the diary, so the true difference could be larger.

For yard-tap households the diaries were about 8 % lower than the meter readings. The meter records everything, including leaks and watering of gardens, while the diaries recorded only household uses. Three of the five diary households told me they watered vegetables in the evening. This means the tap figures are not fully comparable with the standpipe figures, which include almost no garden use. Even if I reduce every tap value by 8 %, however, the tap mean would be about 57 litres, still more than twice the standpipe mean, so the argument holds.

A further limitation is that I measured distance but not time. The JMP's definition of basic access uses a 30-minute round trip including queuing (JMP, n.d.). On two mornings I timed the wait at three standpipes between 6:00 and 7:00. At the standpipes at the foot of T1 and T3, which serve the most households on the village edge, the average wait was about 17 minutes, while at the standpipe at the foot of T4 it was about 3 minutes. These are only two mornings, but they suggest that the households furthest away may also face the longest queues. Time may therefore be a better measure of the "friction" of collecting water than distance, and it could explain why my results differ from the plateau model. If queues are longest at the far standpipes, part of what I have called the effect of distance may really be an effect of waiting.

Correlation does not prove that distance causes lower water use. In Metsimotlhabe, the households furthest from standpipes are mostly on the newer plots at the edge of the village, so distance is partly a measure of how recently a household arrived. Newer residents may have smaller incomes, fewer containers or no wheelbarrow, and any of these could reduce water use. My questionnaire did not ask about these, so I cannot separate them from distance. The fact that household size is not related to distance is reassuring, but it does not rule out the other possibilities.

The secondary sources also have limits. Howard and Bartram's service levels are a synthesis of studies from many countries, so they give a general benchmark rather than a prediction for Botswana. The Drawers of Water studies are detailed and long-term, but they come from East Africa, where rainfall, prices and housing are different. I used them to compare patterns, not exact values, and the similarity of the tap-to-standpipe ratio is more convincing than any single figure.

The survey also took place in one week in April, and the results might have been different in another season. Finally, 48 households is a small sample for one part of one village. The systematic sampling along transects reduced bias, but I had to approach 63 households to obtain 48 responses (a response rate of 76 %). Households that were not at home, perhaps because adults work in Gaborone, are under-represented, and they may use less water during the day than the households I surveyed.

7 Conclusion

In answer to the research question, distance to the nearest water point explains daily water use per person in Metsimotlhabe only to a limited extent. Among households that collect from standpipes, distance has a strong and significant effect (r_s = −0.71), but the effect is modest in size, about 10 litres across the range of distances I measured. Across all households, the type of access is far more important: yard-tap households used 2.7 times as much water as standpipe households, and once type of access is known, distance adds only two percentage points to the variation explained (R² from 0.90 to 0.92). Household size, the cost of water and supply interruptions each explain a little of the remaining variation, but none of them explains the gap between taps and standpipes.

This conclusion is limited by the sample of 48 households, by one week of data and by measuring distance rather than collection time. Even so, the step from standpipe to yard tap is so large and so consistent that it is unlikely to disappear with a larger sample. For a village like Metsimotlhabe, this suggests that adding more standpipes would raise water use only slightly: moving every standpipe household to within 100 m of a standpipe would, on my best-fit line, add only about 9 litres per person for the households now furthest away. Connecting a plot, by contrast, is associated with an increase of almost 40 litres. For planners working towards universal access, connections to the network matter far more than shorter walks, although the extra water used through a tap also raises questions about demand in a semi-arid country.

Two questions remain unresolved. Why do some households stay unconnected when they can see the difference a tap makes, and is the connection fee really the main barrier, or is it the monthly bill that follows? A study comparing recently connected households before and after connection would answer this better than my one-week survey.

Reflective statement — submitted on the Reflection and Progress Form, not counted in the 4,000 words

The most important decision in my research was to let my data change my argument. I started with a simple hypothesis: the further people walk, the less water they use. At the interim reflection session my data already showed that yard-tap households used almost three times as much water as anyone else, and my first reaction was to treat this as a problem and leave the tap households out. Talking it through with my supervisor helped me see that an unexpected result is still a result. I rewrote my argument so that distance became one factor to weigh against the type of access, and the essay became far more interesting because it could answer "to what extent" honestly.

I also learned that a good method is designed, not just carried out. My first sampling plan was to ask my aunt's neighbours. Changing to transects took longer, but in the viva voce I realised that it was the strongest part of my method, because it gave me a full range of distances and a fair mix of older and newer plots. In contrast, I measured distance as a straight line because it was quick. If I did the research again I would measure walking routes and queuing time, because the viva made me see that time, not distance, may be the real "friction" for people carrying water.

In April I visited the village on six days and collected 48 questionnaires, and then I collected the diaries the following week. The diaries taught me to check one source of data against another. When the meter readings and diaries for tap households did not agree, I first assumed I had made an error, but asking the households showed me that the meters were recording garden watering. Investigating a discrepancy taught me more than a perfect match would have done. The same habit helped with my statistics: a coefficient of −0.71 first looked like proof that distance was the answer, and only when I compared it with the size of the effect did I understand what it really showed.

In future research, whether in my university geography course or in any project that relies on surveys, I will run a pilot first. One trial week would have shown me the garden-watering issue and the long queues before my main survey, and I could have measured them properly. I now see research less as finding an answer and more as testing my own assumptions, and that is the skill I most want to take forward.