In this video, I will discuss a more detailed case one in our topology of information treatments. You will recall that this is the case in which households are provided with short term objective information that they may not know. The real disease is typically concentrated in poorly educated populations. Therefore, seems plausible that the one-time provision of objective information about their drinking water quality would help them make better decisions. For example, they could select better drinking water sources or treat their drinking water. Such households may have limited access to information about the health risks of contaminated drinking water, and may not know the quality of their drinking water. They may thus, find it challenging to understand and interpret the publicly available information they received. There's a relatively large literature that examines how this type of policy intervention affects household behavior. I will not try to systematically review this literature in this video. However, my reading of this literature on information treatments in the wash sector is that the estimated effects of one time interventions on desired behaviors and outcomes is smaller than many wash professionals hoped. And that the estimated effects declined rather quickly over time. On the other hand, the cost of information treatment may be low depending on how it is done. So a careful weighing of the costs and benefits of the intervention may still suggest that it is worthwhile. But I think anyone who reads this literature will come to the conclusion that short term provision of objective information is not a silver bullet for solving problems in the wash sector. Instead of attempting to review the many studies on information treatments on the wash sector, in this video, I will discuss one recent paper that I think illustrates well many of the issues, and is representative of the findings from this literature. The paper I will discuss is entitled Information on Household Behaviors to Improve Water Quality By Joe Brown and his co-authors. This paper is one of your readings for this session. We've also included several other papers on this type of information treatment in the wash sector so you can read more on this subject. In this paper, Joe Brown and his co-authors, investigate the question of how households will respond if they're provided with new household specific information about the quality of their drinking water. If households are informed that their drinking water is contaminated, they make take actions, such as the purchase and use of point of use treatment technology to reduce health risks. On the other hand, if a household is informed that their drinking water is not contaminated, it may reduce or stop actions it has already been taking to reduce the risk of contaminated drinking water. This type of research has recently been facilitated by a new drinking water quality test that checks for the presence of hydrogen sulfide. Hydrogen sulfide is a by-product of the metallic processes of some fecal indicator bacteria in water. This drinking water quality test is not perfect. It only yields a binary signal. Either the drinking water sample is contaminated or it is not. The test does not show how contaminated the sample is. Also, most research studies just test one sample from a household. It is quite possible that some samples from a household would be contaminated, and some not. Especially, if households obtain drinking water from multiple sources. This increases the likelihood of positive results that suggest a household's drinking water is worse than it really is. And also have negative results as suggest that the drinking water is safer than it really is. However, because this test is simple and inexpensive, it can be carried out on drinking water samples taken from large number of households. This ability to test the drinking water quality of large numbers of households easily and cheaply. This facilitated the implementation of more rigorous research designs than has been possible in the past. The details of the research design used by Professor Brown and his co-authors are provided in the paper. I will summarize the basic elements of the research design here. Data use for the analysis in this study was collect in three waves. In the summer of 2011, the research team collected baseline information on the socioeconomic characteristics, water use practices, and water quality beliefs of 912 sample households, as well as other information in two peri-urban areas in the outskirts of Phnom Penh, Cambodia. In June and July 2012, they returned and collected drinking water samples from every household in their remaining sample of 848 households. All samples were collected from the main water container that the household used to store its drinking water. For control households, the enumerators took one sample and carried it away when the round one interview was finished, so that the household did not know the results of the test. For treatment households, the enumerator took three samples, all from the same drinking water container. The enumerator carried away the first sample when the interview was finished, so the research team would know the results of the test. The second sample was left with the households, so the members would know the result of the test. And the third sample was treated with an Aquatab disinfection tablet for 30 minute and left with the household. Aquatabs or chlorine point of views household treatment product that disinfects drinking water. The enumerators carry this product with them and made a sales speech to sample households to buy Aquatabs. Enumerators offer the household strips of Aquatab tablets for 25 US cents per strip. Each strip had enough tablets to treat 200 liters of water. That's about two weeks supply of drinking water for most households. So it would cost a household with contaminated drinking water about 50 US cents per month to use Aquatabs to obtain treated, safe drinking water. If the hydrogen sulfide test show that the second untreated sample was contaminated, and a third sample was not contaminated. The household could see that the Aquatab treatment worked as promised to remove bacterial contamination. For the treatment households, the enumerator then stopped the interview to allow time for the hydrogen sulfide test strips to work. The next day, the enumerator returned and checked the test results of the second and third vials that had been left with the treatment households, and discussed the results with the respondent. Then, the enumerators asked questions to assess the respondent's beliefs about the safety of the household's drinking water. These were the same water safety questions that were asked in the baseline survey, before the hydrogen sulfide test. The research team could thus determine if household's beliefs about water safety had changed as a result of the test. Then the enumerator attempted to sell the household strips of Aquatabs. For the Round 2 data collection, the research team returned six weeks later, in August of 2012, to both treatment and control households. The Round 2 questionnaire included the same questions about household health, and water and sanitation behaviors, and water safety beliefs, as in Round 1. As in Round 1, the enumerator attempted to sell the household strips of Aquatabs. In order to assess changes in household's water quality over time, another vial of the household's water was taken for testing in Round 2. If the household had water stored that had been treated by Aquatabs, the enumerator took a sample of this as well. These Round 2 test files were carried away by the enumerators, and the results were not seen by the households. You might wonder why the research team collected drinking water data from all households in the sample, if they were only going to provide half of the sample households with information about the results of the test. There were two reasons. First, the researchers wanted to carefully distinguish between the effect of the information per se in the field where protocol of sampling the drinking water. It is possible, the very process of testing the water could have been a signal to households that the drinking water was contaminated, and the research team wanted to control for this. Second, for ethical reasons, the team wanted to be able to provide information on drinking water quality to the control households at a later period of time. It would have been prohibitively expensive to implement this type of research design that tested drinking water quality of the control households without a simple cheap water quality test. In both round one and round two, the research team collected information on outcome variables. The outcome variables measured by the research team include the perceive safety of drinking water quality, the extent of handwashing, purchase of Aquatabs, use of Aquatabs, self-reported diarrhea. These are the variables that the team hypothesized would change as a result of the provision of information about drinking water quality. Before I summarize Brown's results, it is important to understand the differences between the two peri-urban areas in which the research was conducted. Households in Peri-urban Area 1 were somewhat richer than those in Area 2. And a higher percentage had a pipe water connection. 14% of households in Peri-urban Area 1 had a private piped water connection compared to 1% in Peri-urban Area 2. Households in Area 1 thought that their drinking water quality at baseline was slightly better than in Area 2. And their self-reported willingness to pay for drinking water quality improvements was lower. It is important to emphasize that the research team knew about these differences before the field experiment because they had conducted the round zero baseline survey in 2011. These two peri-urban areas were thus purposely selected for the experiment because they were different. The research team wanted to know how the affects of the information treatment differed in the two studied sites. Remember, randomly selected households in both areas were provided with the results of their drinking water quality tests. Now, let's discuss the results of the information treatment experiment. The results showed that there were only 82 households with uncontaminated drinking water in area one, and 28 households in area two. The research team emphasized that the results for this sub sample of households with uncontaminated water are not as strong, or well powered as for households with contaminated drinking water baseline. I will thus focus on the results of the households with contaminated water baseline. And we're provided with the results of the hydrogen sulfide test. And inform them that their drinking water was contaminated. I will present the results as answers to six questions. Question 1, did the information treatment change households beliefs about the safety of their drinking water? Answer, it did. The results showed that the providing households with information about the water quality of their drinking water, affected their subjective beliefs about health risks. However, this affect was over four times larger in Study Area 1, the poor peri-urban site, and non-statistically significant from zero in Study Area 1. On a ten point scale, on average, treated households in Area 2 had about a one point decrease. Treated households in Study Area 1 reported only a zero point, two point decrease. Question 2, did the information treatment increase households' purchase of Aquatabs? Answer, yes, in Study Area 2. No, in Study Area 1. In Round 1, in Study Area 2, the first time the enumerators tried to sell Aquatabs, 53% of the control households purchased Aquatabs, and 65% of the treatment households purchased Aquatabs. This difference of 12% was statistically significant. In Round 1 in Study Area 1, 47% of the control households purchased Aquatabs, and 50% of the treatment households purchased Aquatabs. This difference of 3% was not statistically significant. However, by Round 2, the effect of the information treatment had no affect on the purchase of additional Aquatabs. This suggests that the initial positive affect of the information treatment in Study Area 2 disappeared after only six weeks. Question 3, did the information treatment increase households' use of Aquatabs? We know that many households in both the treatment and controls purchased Aquatabs, but did they actually use them? Answer, of those households that purchased Aquatabs in Round 1, 90% bought six or fewer strips which is 60 or fewer tabs. If those households use the Aquatabs as recommended for all their drinking water between the first and second rounds, almost all households should have used up their supply of Aquatabs. But the research team's result show that many households still had Aquatabs on hand. Of those households that still had their Aquatabs packing in Round 2. Only about a third had used more than half of the Aquatabs they'd purchased in Round 1. This finding means that households were not using the Aquatabs all the time as recommended. Two-thirds of households who purchased Aquatabs in Round 1 used them less intensively than recommendations would imply. Nearly 80% of purchasers reported not currently using them on the day of Round 2 interview. These results suggest that the information treatment is unlikely to result in the desired health outcomes. Question 4, did the information treatment increase household's hand washing? Answer, yes, in Study Area 1. No, in Study Area 2. Around to 20% of the control households in Study Area 1 reported handwashing compared to 33% of the treatment households. This difference of 13% was statistically significant. However, in Study Area 2, 18% of the control households reported handwashing compared to 20% of the treatment households. This difference of 2% was not statistically significant. Question 5, did the information treatment increase the quality of households' drinking water? Answer, no. In Round 2, there was no significant difference in the drinking water quality between treatment and control households. The provision of information about contaminated drinking water did not result in measured drinking water quality. Question 6, did the information treatment reduce households' self-reported diarrhea? In Round 2, there was no significant difference between self-reported diarrhea among treatment and control households. So to summarize, most households appeared to believe the findings of this hydrogen sulfide test, when it showed that they're drinking water was contaminated. These households reported that their drinking water was less safe than they initially thought. Over half of the households in study area two purchased Aquatabs. But the difference between the treatment and control households was not large, 52% versus 64%. The information treatment had no effect on Aquatab purchases in study area one. In neither study area were Aquatabs used as recommended, either by the treatment or controlled households. The information treatment had no affect on the repurchase of Aquatabs in round two in either study area. The information treatment seems unlikely to have had much affect on health outcomes because the research shows no affect on drinking water quality or self-reported diarrhea. What are we to make of all this? I think there are two key messages. First, it would be tempting simply to conclude that case one information treatments don't work. But I think the lesson here is more nuanced. I think the lesson is that, it's very difficult to generalize findings from such experiments to other locations. The results differ even between these two peri-urban areas in Phnom Penh. The claim that more evidence is needed to make policy in the wash sector seems tenuous to me. A few more randomized control trials such as this, are not going to do much to eliminate the uncertainty associated with the affects of these kinds of policy interventions. Second, I interpret the results from this research as re-enforcing the case to decentralized decisions about the effectiveness of policy interventions. And program designed to the local and regional level where planners and government officials are more likely to have first hand experiential information about how effective policy interventions will be. This does not mean that evidence such as presented in this paper is unhelpful to local decision makers. But the findings should caution us not to be overly confident about what we know. We should not simply assume that if we find a treatment effective of policy intervention in one study area, that it will be the same elsewhere. Not only may local conditions be different, but also the causal mechanism itself may work differently in different places. In the next video, we will discuss case two of our topology of information treatments.