[BLANK_AUDIO]. The literature on modeling household water demand in developing countries has grown grown quickly in the last decade, but it is still surprisingly thin. In this short video, before we discuss empirical results from the literature on household water demand, I want you to step back for a minute and think about the challenges of modeling household water demand behavior in developing countries. Why is it different than modeling household water demand in industrialized countries? Consider first a household in an industrialized country with 100% coverage with piped water services. This household has one main source of water for all its water uses, drinking cooking, bathing, washing, lawn watering, etc. This means that if we want to model household water use, water can be a homogeneous good. In other words, the attributes of the water used, quality, reliability, and price are the same for all water uses. This is not true in many developing countries, especially in rural areas. A household in many rural communities must choose between several water sources. For a household in a developing country each water source may have a different quality, reliability, and price. Some sources may be cheap, but far away. Others, such as water from water vendors, may be expensive but very convenient. In such a community water is a heterogeneous good. Households may use different water sources for different purposes, or they may use multiple sources for the same use. The main point is that in an industrialized country a household hardly has to think about which water source to use. The only exception might be whether or not they decide to buy bottled water to drink. In the literature on household water demand modeling in industrialized countries the household's choice of water source is not a major modeling issue, because it can be simply assumed that a household will connect to a piped water network if it's available. In a developing country this is not the case. Even when a piped water connection is available a household may choose not to connect because the quality of water from the tap is so poor, or because the service is so unreliable. A household in a developing country must decide how to match different water sources with different water uses. What this means is that the first step in a household water demand model must be to understand the household's choice of water source. In an industrialized country an analyst trying to model household water demand will try to obtain actual data on water usage and price from customer billing records from the water utility. If she is unable to get these data she will usually go to another utility or find another research topic. Analysts rarely even contemplate collecting primary data from households about water use. However, in a rural community in a developing country the analyst will have to collect primary data. General purpose household surveys conducted in developing countries do have a few questions on household water use, but such surveys never have the data you would need to carefully model household decisions on water source choices. This is because one needs data on not only the source or sources chosen by the household, but also on all the attributes of the sources not chosen. It is a difficult task to collect accurate data on water source choices and water use. Even in urban areas where utilities do exist, working with the utilities customer billing records in developing countries presents different challenges than customer data from utilities in industrialized countries. First, customer billing records from a utility in a developing country may not be computerized. It would then be a huge task to get data into computer readable files. Second, it is not always the case that customer billing records have actually used the official water tariff to calculate water bills. In this case, determining the price charged to customers can be very challenging. Third, many households, especially in large apartment blocks, may not have metered connections. This is not a problem unique to developing countries. But the scale and challenge that unmetered connections pose is much larger in many cities in developing countries. Fourth, not all households in a city will have a piped water connection from the utility. In this case, the analyst can obtain only a partial picture of household demand by examining the utility's customer billing records. Fifth, another challenge is that household data obviously will not be avail for levels of service that do not yet exist in the community. The challenge of understanding how people will respond to new levels of water services is especially difficult. This is because people may not know their preferences for technologies they have not seen or used. It's particularly true for unfamiliar sanitation technologies such as condominial sewers or composting toilets. The Hawthorne effect is named for a famous experiment that was conducted in the United States. In the experiment the researcher found that subjects improve an aspect of their behavior that is being measured by the researcher, simply in response to the fact that they are being studied. In other words, the researcher did not introduce any particular experimental manipulation, like giving out new information to respondents. Just the fact that the subjects were being watched changed their behavior for the better. The Hawthorne Effect could easily occur in villages where researchers are watching how households use water. Another way to think about this problem is that watching itself is a type of intervention. This is related to the problem of interviewer bias in household surveys. This is when respondents tell researchers what they think the researchers want to hear. The Hawthorne Effect and in, interviewer bias are both added complications in collecting data needed to model household water use in developing countries, particularly in rural areas where there may be cultural and socioeconomic differences between the visiting research team and people living in the community. To wrap up, the results from the empirical studies we will look at have involved challenging data collection problems. [BLANK_AUDIO]
