Cambodia Sanitation Marketing Scale-Up

Transcription

1 Cambodia Sanitation Marketing Scale-Up SMSU 2.0 M&E Mid-term Results Rachel Rose - Research and Evaluation Specialist Veasna Tour - ide Cambodia SMSU M&E Manager

2 Cambodia SMSU 2.0 Mid-term Results Page 2 CONTENTS INTRODUCTION... 4 METHODS... 4 Overview of methods... 4 Latrine Count... 4 Sales Tracking... 5 Customer Survey... 6 Latrine Business Profile Update... 6 COVERAGE... 6 SMSU Milestones and Survey Findings... 6 Coverage by Province... 8 Coverage by District... 9 Abandonment Sharing Diarrhea Rates SALES Non-project-connected sales Purchases by poor households INSTALLATION CALL CENTER RESULTS Confirmed cancellations Acquired latrine Wrongful cancellations Summary findings DMQ RESULTS Primary objection Financial barrier Other barriers BUSINESS ANALYSIS Summary of SMSU latrine business operators (LBOs) Gender analysis... 25

3 Cambodia SMSU 2.0 Mid-term Results Page 3 TABLES Table 1: Latrine count sample profile... 5 Table 2: Proportion of households sharing a latrine comparison Table 3: Non-project-connected ratio comparison Table 4: Non-project connected ratio, SMSU 2.0 mid-term update Aug Table 5: IDPoor sales, by province Table 6: Below ground and superstructure installation rates for SMSU 1.0 and Table 7: Customer calls, by province Table 8: Call center classification groups Table 9: Re-classification of call center groups Table 10: Snapshot of SMSU 2.0 LBOs FIGURES Figure 1: Coverage rates comparison, across all SMSU Provinces... 7 Figure 2: Coverage rates over time, across all SMSU provinces... 7 Figure 3: Wet coverage rates over time, by province... 8 Figure 4: Wet latrine coverage update, by province... 9 Figure 5: IDPoor latrine coverage update, by province... 9 Figure 6: Change in wet latrine coverage, by district Figure 7: Change in IDPoor latrine coverage, by district Figure 8: Latrine coverage and diarrhea rates correlation Figure 9: Superstructure installation curves for SMSU 1.0 and Figure 10: SMSU 2.0 superstructure installation rates by IDPoor Figure 11: Below ground installation rate, by IDPoor Figure 12: Superstructure wall materials, by IDPoor Figure 13: Reason for cancellation Figure 14: Primary objection to latrine purchase, by province Figure 15: Primary objection to latrine purchase, by IDPoor Figure 16: Primary financial barrier, by province Figure 17: Price composition of latrine set from LBO (includes profit)... 25

4 Cambodia SMSU 2.0 Mid-term Results Page 4 INTRODUCTION This report summarizes findings from ide s monitoring and evaluation efforts throughout the Sanitation Marketing Scale-Up (SMSU) 2.0 program implemented in Cambodia from November 2014 to December This report is an accompaniment to the SMSU 1.0 Monitoring and Evaluation Results Update and will reference progress made against inception of SMSU 2.0 (the endline of SMSU 1.0) as well as progress against all SMSU project time (since September 2011). These results inform ongoing program decision-making and provide the basis for determining whether the project met the initial key targets set under SMSU 1.0 as well as additional targets established with SMSU 2.0, in partnership with Stone Family Foundation, Department of Foreign Affairs and Trade Australia, and TripAdvisor. The report outlines the methods used to collect and analyze data and provides the key findings related to each of the project targets, organized around the themes of coverage and use, sales and business development. METHODS It is important to note that some data collection methods are continuous throughout program operations. Any estimates, figures, or data calculated from continuous data collection (such as customer surveys and DMQ surveys) are inclusive of results up to September Other data collection efforts, such as the latrine count, call center analysis, and gender analysis are static collection efforts. Date ranges for those data are noted in the report. Overview of methods The key indicators of success for this project are: Sales through project-connected enterprises Sales to poor households through project-connected enterprises Sales through non-project-connected enterprises in project areas Sales through non-project-connected enterprises outside project areas Increase in latrine coverage (and associated diarrhea rates as self-reported) Consistent use by adults (self-reported) In order to measure these results, we have used the following methods: Monthly collection of sales data from project-connected enterprises (measures direct sales) Ongoing verification of latrine-business sales (including a household questionnaire to identify the percentage of poor customers, consistent use by adults and installation rate) An biennial latrine count to estimate changes in coverage at district level and calculate sales through other non-project-connected enterprises in project areas. A quarterly business profile update, collecting key business data from each active latrine business. o Including a one-time gender analysis studying the impact of female involvement in latrine businesses A one-time call center analysis to follow-up on canceled orders Ongoing data survey for all sales pitches that do not result in an order to help determine barriers to purchase Latrine Count ide conducted the first interim latrine count for SMSU 2.0 between June and September 2016, using a cluster sample method, following up on a subset of 15 villages within each district of the seven SMSU provinces. Within

5 Cambodia SMSU 2.0 Mid-term Results Page 5 each cluster (village) we conducted a census of households to establish the total latrine count. ide research assistants (RA) worked with local leadership to draw a sketch map of the village, identifying the location of all households. The RA visited each household to record the following data: IDPoor status (1/2/none) latrine type (wet/dry/none) installed (y/n) whether the latrine was subsidized (y/n) if no latrine, whether households shared a latrine with anyone else diarrhea prevalence While surveys are administered at a household, the results are aggregated to the village level and our 205,318 household observations are collapsed to 855 primary sampling units for analysis. Table 1 presents the sample profile for the most recent round of latrine count conducted during SMSU 2.0 T ABLE 1: LATRINE COUNT SAMPLE PROFILE PROVINCE DISTRICTS VILLAGES HOUSEHOLDS Banteay Meanchey ,377 Kampong Thom ,291 Kandal ,630 Oddar Meanchey ,874 Prey Veng ,528 Siem Reap ,038 Svay Rieng ,580 Total ,318 RAs collected household data using paper forms, which were field checked and further validated by the M&E manager at ide s head office in Phnom Penh. Data entry clerks entered the data at the Phnom Penh office. The SMSU M&E coordinator conducted field verification of latrine count data. Sales Tracking ide s team of research assistants collects sales data from each project-connected latrine business every month. Latrine businesses maintain a record book designed by ide to track customer data (name, phone number, and village/commune/district/province) and details of the sale (units purchased, components, sales agent responsible, etc.). Research assistants record summary data on deliveries based on the record book and also take a digital photo which they send to headquarters for verification purposes. Verification of monthly delivery data records ensures that deliveries are accurately and properly recorded in LBOs book. Quick verification is implemented on a monthly basis and through the following process: 1. LBOs with greater than 35 reported deliveries in a month are stratified by province. The M&E Coordinator selects LBOs from each province by using probability proportional to size sampling, replacing those who were randomly selected the previous month. 2. Research assistants are assigned two LBOs for verification. 10 customers are randomly selected per LBO from his or her record book for follow-up. 3. Research assistants complete field verification report and sync to Salesforce database. If there are at

6 Cambodia SMSU 2.0 Mid-term Results Page 6 minimum two misreported cases per LBO, the M&E team conducts follow-up checks for all deliveries of the LBO during the month. In this report, non-project connected sales figures are calculated based on the latrine count. Thus, sales figures presented in Table 4 only represent sales from November 1, 2014 to Aug 31, Estimates of sales to poor households in Table 5 include are calculated from customer surveys, and thus include data collected through September Customer Survey Follow-up surveys are given to latrine customers to learn about their poverty status, expenses for latrine, use, and satisfaction. Customer surveys also serves as a tool of delivery data verification during the period. Sampling procedures for the customer survey are as follows: 1. The sampling frame is restricted to customers whose latrine was delivered 6 to 12 months prior. LBOs with the highest delivery counts are selected proportional to the research assistants covering that province customers are randomly selected per LBO, with additional customers selected for replacement as needed during field work. 3. Research assistants are provides with a list of selected customers and their contact information via the mobile app, TaroWorks. Interviews are to alternated between male head of household and female head of household (or spouse). 4. Completed surveys are synced to Salesforce database. The M&E coordinator regularly checks data to ensure results are complete and accurate. Latrine Business Profile Update Every quarter, RAs collected a set of business data from latrine businesses. These data include information on staffing, pricing, inventory, capital investments and financing. The purpose of this exercise is to build up a picture over time of the businesses providing toilets, provide a descriptive analysis of market dynamics and identify determinants of business success or failure. Data from this exercise, combined with sales data, support the business analysis in this report. Latrine business operations data in this report are from the time period November 2014 to December COVERAGE SMSU Milestones and Survey Findings ide has conducted five comprehensive household surveys to estimate latrine coverage rates in the seven program provinces. The most recent latrine count under SMSU 2.0 was completed in August Figures 1 7 present the changes in coverage rates over time and by latrine type. A note about terminology: Wet refers to installed improved pour-flush latrines Poor refers to poor households (identified by IDPoor1 or IDPoor2 status) with installed wet latrines Installed refers to both wet and dry latrines that are installed. In Figure 1 we see statistically significant increases in coverage rates across all three metrics between the start of SMSU 1.0, end of SMSU 1.0, and SMSU 2.0 mid-term.

7 Cambodia SMSU 2.0 Mid-term Results Page 7 F IGURE 1: COVERAGE RATES COMPARISON, ACROSS ALL SMSU PROVINCES 12% 24% 34% 29% 45% 56% 30% 46% 57% Poor households with latrine Households with wet latrine Households with installed latrine Baseline Feb End of SMSU 1.0 Aug SMSU 2.0 Mid-term Aug Figure 2 presents the latrine coverage rates for all rounds of data collection completed since the start of project. Note that latrine count frequency changed from every year to every two years under SMSU 2.0. Due to this change in frequency, 2015 coverage rates were calculated to create equal visual spacing and thus representation on all line graphs. 70% 60% 50% 40% 30% 20% 10% 0% F IGURE 2: COVERAGE RATES OVER TIME, ACROSS all SMSU PROVINCES Feb Aug Aug Aug (imputed) Aug Installed 30% 33% 38% 46% 51% 57% Wet 29% 32% 37% 45% 50% 56% Poor 12% 10% 16% 24% 29% 34%

8 Cambodia SMSU 2.0 Mid-term Results Page 8 Coverage by Province Figures 3, 4, and 5 present latrine coverage rates by province over time. We see that Kandal province consistently has the highest latrine coverage rates while Oddar Meanchey has the lowest. This data supports field reporting from Cambodia staff about the challenging environments and market conditions in Oddar Meanchey. 70% 60% 50% 40% 30% F IGURE 3: W ET COVERAGE RATES OVER TIME, BY PROVINCE Kandal Svay Rieng Banteay Meanchey Prey Veng Siem Reap Kampong Thom Oddar Meanchey 20% 10% 0% Feb Aug Aug Aug (imputed) Aug Kandal 49% 50% 57% 62% 66% 69% Svay Rieng 29% 26% 37% 46% 55% 65% Banteay Meanchey 33% 34% 42% 48% 53% 57% Prey Veng 19% 25% 28% 39% 46% 53% Siem Reap 17% 31% 28% 35% 41% 48% Kampong Thom 24% 18% 31% 37% 42% 46% Oddar Meanchey 17% 23% 22% 26% 30% 35% Figures 4 and 5 show the provincial change in wet latrine coverage and poor latrine coverage, respectively. To note, most provinces, but not all, have shown statistically significant increases in latrine coverages rates between the end of SMSU 1.0 and this mid-term update as noted by the error bars. 1 All provinces have a statistically significant increase in coverage rates between the start of SMSU 1.0 and this mid-term update. 1 Error bars represent the 95% confidence interval for each coverage rate estimate.

9 Cambodia SMSU 2.0 Mid-term Results Page 9 F IGURE 4: W ET LATRINE COVERAGE UPDATE, BY PROVINCE 33% 48% 57% 24% 37% 46% 49% 62% 69% 17% 26% 35% 19% 39% 53% 17% 35% 48% 29% 46% 65% 29% 45% 56% Banteay Meanchey Kampong Thom Kandal Oddar Meanchey Prey Veng Siem Reap Svay Rieng ALL Baseline Feb End of SMSU 1.0 Aug SMSU 2.0 Mid-term Aug F IGURE 5: IDPOOR LATRINE COVERAGE UPDATE, BY PROVINCE 12% 26% 38% 13% 20% 28% 15% 36% 42% 11% 9% 17% 5% 23% 33% 7% 14% 26% 24% 30% 47% Banteay Meanchey Kampong Thom Kandal Oddar Meanchey Prey Veng Siem Reap Svay Rieng Baseline Feb End of SMSU 1.0 Aug SMSU 2.0 Mid-term Aug Coverage by District Figures 6 and 7 present the growth in latrine coverage rates at the district level for both wet and poor latrine ownership. The percentage values listed are the rate of latrine coverage estimated at the mid-term update. Within each province group, it is easy to identify which districts are driving down the rates of latrine coverage.

10 Cambodia SMSU 2.0 Mid-term Results Page 10 F IGURE 6: CHANGE IN WET LATRINE COVERAGE, BY DISTRICT Otdar Meanchey Kandal Kampong Thom Banteay Meanchey Svay Rieng Siem Reap Prey Veng Mongkol Borei Preah Netr Preah Ou Chrov Malai Phnum Srok Svay Chek Thma Puok Baray Stoung Kampong Svay Prasat Balangk Prasat Sambour Santuk Sandan Angk Snuol Kien Svay Mukh Kampul Khsach Kandal Kandal Stueng Leuk Daek Lvea Aem Ponhea Lueu Kaoh Thum Chong Kal Anlong Veaeng Banteay Ampil Trapeang Prasat Ba Phnum Kampong Trabaek Pea Reang Sithor Kandal Peam Ro Kamchay Mear Svay Antor Preah Sdach Peam Chor Me Sang Kanhchriech Puok Soutr Nikom Chi Kraeng Srei Snam Prasat Bakong Kralanh Angkor Chum Banteay Srei Angkor Thum Varin Svay Leu Rumduol Svay Teab Svay Chrum Chantrea Kampong Rou Romeas Haek 23% 70% 63% 59% 50% 50% 48% 39% 53% 50% 45% 42% 41% 38% 34% 80% 76% 75% 74% 74% 71% 64% 59% 48% 43% 38% 32% 29% 65% 61% 59% 58% 57% 55% 49% 48% 46% 46% 43% 55% 55% 53% 49% 46% 46% 40% 40% 35% 33% 69% 68% 65% 64% 64% 63% Baseline Feb End of SMSU 1.0 Aug SMSU 2.0 Mid-term Aug. 2016

11 Cambodia SMSU 2.0 Mid-term Results Page 11 F IGURE 7: CHANGE IN IDPOOR LATRINE COVERAGE, BY DISTRICT Otdar Meanchey Kandal Kampong Thom Banteay Meanchey Svay Rieng Siem Reap Prey Veng Preah Netr Preah Mongkol Borei Ou Chrov Phnum Srok Svay Chek Malai Thma Puok Prasat Sambour Stoung Baray Prasat Balangk Kampong Svay Sandan Santuk Kien Svay Angk Snuol Khsach Kandal Mukh Kampul Kandal Stueng Leuk Daek Lvea Aem Ponhea Lueu Kaoh Thum Chong Kal Banteay Ampil Anlong Veaeng Trapeang Prasat Kampong Trabaek Ba Phnum Peam Ro Kanhchriech Sithor Kandal Kamchay Mear Pea Reang Svay Antor Me Sang Preah Sdach Peam Chor Soutr Nikom Srei Snam Chi Kraeng Puok Kralanh Banteay Srei Angkor Chum Angkor Thum Varin Prasat Bakong Svay Leu Rumduol Romeas Haek Svay Chrum Svay Teab Chantrea Kampong Rou 6% 9% 15% 17% 23% 21% 24% 27% 26% 25% 21% 21% 21% 18% 34% 31% 29% 26% 28% 27% 27% 25% 23% 31% 36% 36% 35% 32% 31% 31% 30% 29% 33% 32% 32% 26% 36% 39% 45% 44% 42% 41% 40% 41% 47% 44% 43% 44% 51% 50% 50% 50% 48% 47% 45% Baseline Feb End of SMSU 1.0 Aug SMSU 2.0 Midterm Aug. 2016

12 Cambodia SMSU 2.0 Mid-term Results Page 12 Abandonment The rate of abandoned latrines remains constant at 0.3% at the time of this most recent latrine count. At the end of SMSU 1.0 the abandoned rate was 0.2% Sharing Table 2 presents the proportion of households sharing a latrine among non-latrine owners and within the larger population. Among households that do not own a latrine, the percent of those sharing a latrine with another household has increased from 8% to 15%. This suggests that in addition to improved wet latrine coverages rates increasing, more households are also gaining access to a latrine, though via sharing. This could indicate more households entering the sanitation ladder and more inclined for the next level of latrine ownership. T ABLE 2: PROPORTION OF HOUSEHOLDS SHARING A LATRINE COMPARISON % Sharing of Non-latrine owners % Sharing of Population End of SMSU 1.0 Aug SMSU 2.0 Mid-term Aug End of SMSU 1.0 Aug SMSU 2.0 Mid-term Aug Kandal 19% 31% 7% 9% Svay Rieng 11% 22% 6% 7% Kampong Thom 8% 16% 5% 8% Banteay Meanchey 6% 14% 3% 6% Prey Veng 6% 13% 4% 6% Oddar Meanchey 1% 8% 1% 5% Siem Reap 3% 7% 2% 4% ALL 8% 15% 4% 7% Diarrhea Rates The latest round of the latrine count collected data on diarrhea incidence within the household with the goal of estimating the relationship between latrine coverage and diarrhea rates. Preliminary regression analysis tested for any statistically significant relationship between latrine coverage and diarrhea rates. Our regression model uses the rate of diarrhea incidence as the dependent variable, the percentage of households with latrines as the independent variable, and several key control variables (percent coverage change, percent households with electricity by province, percent of households with subsidized latrine installations, percent of households with a shared latrine, percent of ID-poor households, and district fixed effects). The results from our model show that a 1 percentage point increase in households with latrine coverage during the August 2016 survey is linearly associated with a statistically significant decrease of 0.03 percentage points in the rate of diarrhea incidence. Our conclusion is that increases in latrine coverage rates is negatively and significantly correlated with diarrhea rates, though the magnitude of the change is quite small. Further, we found that district level factors (which we interpret as potential differences in infrastructure, weather patterns, sanitation concentration) clearly have an impact on the rate of diarrhea incidences throughout Cambodia. Figure 8 plots the district level relationship between diarrhea rates and coverage rates, grouped by province. Though our initial analysis finds a statistically significant but small in magnitude relationship between latrine coverage and diarrhea rates, we are encouraged by the findings. More sophisticated modeling and series of robustness checks are forthcoming to better understand the impact of latrine coverage on diarrhea rates in a community. This includes the adding another round of latrine count data, differentiating between diarrhea rates for children under five and adults, and assessing different threshold levels of coverage. It is important to recall that we are measuring this relationship at a village level not at the household, thus measuring the spillover effect of latrine coverage. That is to say, though a household may purchase a latrine and decrease their incidence of diarrhea within their respective household, other households in the area may be practicing poor hygiene and have high incidence. This analysis is attempting to demonstrate that improving latrine coverage at a community level, not just household, has a positive health impact for the community. We would expect a study of household level effects to be much larger in magnitude.

13 Cambodia SMSU 2.0 Mid-term Results Page 13 35% F IGURE 8: LATRINE COVERAGE AND DIARRHEA RATES CORRELATION 30% Two Week Diarrhea Prevalence Rate 25% 20% 15% 10% 5% SALES 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% Non-project-connected sales Improved Latrine Coverage Rate Banteay Meanchey Kampong Thom Kandal Otdar Meanchey Prey Veng Siem Reap Svay Rieng From Table 3 we see that the non-project-connected ratio at the time of SMSU 2.0 mid-term update is 0.7 compared to 1.1 at the end of SMSU 1.0, indicating that SMSU is making up a greater share of the market. In both instances, Kandal is the province with the highest ratio of project connected to non-project connected sales. That is, for roughly every one SMSU latrine sold in Kandal approximately 3.3 other latrines are sold. Under SMSU 1.0 Kandal s ratio was five to one. We see ratio declines in all other provinces, except for Oddar Meanchey, where previously the ratio was 0.4 and it is now 2.7, indicating a rise in non-smsu latrine providers in that province. It is important to note that the program was not active in the Oddar Meanchey for the first years of this reporting period, which may explain lagging results. T ABLE 3: N ON-PROJECT-CONNECTED RATIO COMPARISON End of SMSU 1.0 Ratio SMSU 2.0 Update Ratio Banteay Meanchey Kampong Thom Kandal Oddar Meanchey Prey Veng Siem Reap Svay Rieng ALL PROVINCES Table 4 shows the raw counts and ratios of SMSU installed latrines as compared to non-project connected latrines. This data is inclusive of sales records only from November 2014 to August 2016, as the ratio calculation is dependent upon the latest latrine count.

14 Cambodia SMSU 2.0 Mid-term Results Page 14 T ABLE 4: N ON- PROJECT CONNECTED RATIO, SMSU 2.0 MID- TERM UPDATE AUG New Latrines SMSU Latrines Non-SMSU Latrines Ratio Banteay Meanchey 12,979 8,604 4, Kampong Thom 17,819 17, Kandal 22,723 5,309 17, Oddar Meanchey 5,579 1,522 4, Prey Veng 48,587 26,856 21, Siem Reap 25,120 14,935 10, Svay Rieng 26,381 16,588 9, ALL PROVINCES 157,813 91,208 66, Purchases by poor households Of households interviewed, we find that on average 21% of sales are to poor customers (classified either IDPoor1 or IDPoor2). The project average is the same as the rate of poor sales at the end of SMSU 1.0 though we find provincial variations. For example, rates of poor sales in Oddar Meanchey and Siem Reap decline while Svay Rieng rates improved. Overall there is now a more equal distribution of poor sales across all provinces. T ABLE 5: IDPOOR SALES, BY PROVINCE SMSU 1.0 End of project SMSU 2.0 Mid-term update Oddar Meanchey 30% 25% Prey Veng 23% 25% Kandal 24% 24% Banteay Meanchey 23% 22% Kampong Thom 23% 21% Svay Rieng 8% 21% Siem Reap 22% 15% ALL PROVINCES 22% 21% INSTALLATION For SMSU 1.0 and the start of SMSU 2.0, our method of calculating installation rates was actually an approximation of usage. When our monitoring team surveyed a household, an installed latrine was recorded if the household had below ground installation (latrine pit rings and slab) and an installed superstructure. The working assumption was that a latrine was only fully installed once it contained both the substructure and a superstructure for privacy as a latrine without a superstructure was unlikely to be used. Our customer survey data also supported this theory as 98% of all households with installed latrines reported adults using them always or almost always. However, we saw that following this definition, installation rates remained a consistent 62-65% between SMSU 1.0 and the mid-term of SMSU 2.0. This calculation of installation does not accurately capture the change in program implementation when the latrine price increased to include below ground installation. 2 To better understand these rates, we separated our survey question on installation to ask about below ground installation and superstructure installation independently. Table 6 presents the installation rates for superstructure installation and below ground installation under SMSU 2.0 in comparison to SMSU 1.0 superstructure installation rates. This data is inclusive of all customer surveys completed through September Starting March 2015, latrine purchase price is inclusive of and limited to substructure installation.

15 Cambodia SMSU 2.0 Mid-term Results Page 15 T ABLE 6: BELOW GROUND AND SUPERSTRUCTURE INSTALLATION RATES FOR SMSU 1.0 AND 2.0 SMSU 1.0 Superstructure rate SMSU 2.0 Superstructure rate SMSU 2.0 Below ground rate Banteay Meanchey 72% 71% 96% Kampong Thom 48% 48% 86% Kandal 64% 75% 94% Oddar Meanchey 41% 49% 72% Prey Veng 70% 67% 78% Siem Reap 54% 62% 98% Svay Rieng 68% 62% 86% ALL PROVINCES 62% 65% 89% We see that while overall installation rates, inclusive of superstructure and below ground structures, are marginally better, below ground installation rates average 89% for the project region. Given the change in the sales product (i.e. below ground installation included with latrine purchase) we would expect this number to be close to 100% and do see 90% or higher installation rates in three provinces, Banteay Meachey, Kandal, and Siem Reap. Interestingly, Oddar Meanchey and Prey Veng both have below ground installation rates in the seventies. Further research should be given to determine what is driving those lower installation rates. In addition to the overall installation rates presented in Table 6, we can estimate how long households take to install superstructures after purchase of a latrine. Figure 9 presents the installation curves for SMSU 1.0 and 2.0. To note we have two curves for SMSU 2.0, the first is the installation rate of superstructures on all latrines delivered under SMSU 2.0 (whether installed below ground or not). The second is the installation rate of superstructure on latrines that already have their below ground structure installed. We see that within one year 73% of households will install a superstructure if they have their below ground structure installed. Notably, half of all households will have a superstructure installed within one month of purchase if they have their below ground structure installed. F IGURE 9: S UPERSTRUCTURE INSTALLATION CURVES FOR SMSU 1.0 AND % 80% 60% 40% 20% 0% For all latrines 0% 36% 49% 52% 54% 55% 56% 58% 61% 63% 64% 66% 67% 2.0 For all latrines 0% 42% 51% 55% 57% 59% 61% 62% 62% 63% 63% 63% 64% 2.0 For installed sub 0% 49% 59% 64% 68% 70% 71% 72% 72% 72% 73% 73% 73% Number of Months Post-Purchase until Installed 3 SMSU 2.0 data is restricted to purchases after March 2015 when purchase is inclusive of and limited to substructure installation

16 Cambodia SMSU 2.0 Mid-term Results Page 16 We are also interested to know if IDPoor households follow the same installation trajectory when compared to non-poor households. Figure 10 shows the superstructure installation curves for households based on their IDPoor classification. Interestingly, we see that IDPoor households have slightly higher superstructure installation rates than non-poor households. Within one month of latrine purchase, 51% of IDPoor 1 households have a superstructure installed, whereas only 41% of non-poor households do (46% of IDPoor 2 household superstructures are installed within one month). F IGURE 10: SMSU 2.0 SUPERSTRUCTURE INSTALLATION RATES BY IDPOOR 4 100% 80% 60% 40% 20% 0% Nonpoor 0% 41% 49% 54% 57% 59% 60% 61% 62% 62% 62% 63% 63% 63% 63% 63% IDPoor1 0% 51% 54% 57% 58% 60% 61% 63% 63% 64% 64% 64% 64% 65% 65% 65% IDPoor2 0% 46% 55% 58% 60% 62% 64% 65% 65% 65% 66% 66% 66% 66% 66% 66% Number of Months Post-Purchase until Installed There are two potential explanations regarding this difference in superstructure installation rate. First, we find different rates of below ground installation among poor and non-poor households as shown in Figure 11. IDPoor1 household have the highest rates of below ground installation (94%), while IDPoor2 have the lowest (84%). The higher rates of IDPoor 1 below ground installation could help explain why IDPoor1 also has the highest superstructure installation rate one month after purchase. However, IDPoor2 s lowest rate of below ground installation challenges this theory. 4 SMSU 2.0 data is restricted to purchases after March 2015 when purchase is inclusive of and limited to substructure installation

17 Cambodia SMSU 2.0 Mid-term Results Page 17 F IGURE 11: BELOW GROUND INSTALLATION RATE, BY IDPOOR Nonpoor 90% 1% 9% IDPoor1 94% 6% IDPoor2 86% 2% 12% Installed Installing Uninstalled Another explanation is that while IDPoor households are quicker to install a superstructure than non-poor households are, the superstructure may be constructed of lower quality or less durable materials. Figure 12 shows us that almost half of all non-poor households (48%) construct their superstructure walls out of concrete or brick, while on 20% of IDPoor1 households do and 32% of IDPoor2 households do. Instead, the primary material for superstructure walls in poor households is galvanized steel (44% IDPoor1 and 43% IDPoor2; compared to 28% non-poor). Finally, we also see higher rates of poor households building their superstructure walls out of bamboo, leaves, or other thatch materials than non-poor households. F IGURE 12: S UPERSTRUCTURE WALL MATERIALS, BY IDPOOR Nonpoor IDPoor1 IDPoor2 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Concrete / brick Galvanized steel Bamboo / thatch Plastic sheet Wood Other CALL CENTER RESULTS The WASH Call Center was established on 17 October 2016, with six interns and one ide Research Assistant. The purpose was to contact customers with cancelled orders in Salesforce, from the period of January 2016 to July The program wanted to learn the reasons causing these customers to cancel and to identify if any customers orders were wrongfully designated as cancelled. By February 2017, the WASH Call Center called 13,040 customers across seven provinces. We successfully contacted 7,227 customers, but 5,813 customers could not be reached. (Note that there are several reasons why a customer is unable to be reached. The most common reasons are that the household are frequently switching operators and the Village Chief does not recognize the customer s name.) Table 7 shows the number of successful and unsuccessful calls made, segmented by province:

18 Cambodia SMSU 2.0 Mid-term Results Page 18 T ABLE 7: CUSTOMER CALLS, BY PROVINCE No. of Customer Calls Successfully Contacted Success Rate Banteay Meanchey % Kampong Thom 3,271 1,669 51% Kandal 2,445 1,512 62% Oddar Meanchey % Prey Veng 2,705 1,294 48% Siem Reap 1,994 1,107 56% Svay Rieng 1, % ALL PROVINCES 13,040 7,227 55% Of the 7,227 successfully contacted customers, 77% confirmed a canceled order, 21% reported acquiring a latrine, and 3% were an erroneous cancelation. Table 8 shows the breakdown of the customer classifications. T ABLE 8: CALL CENTER CLASSIFICATION GROUPS 1. Confirmed Cancellations 5,557 77% 2. Customer reporting having acquired a latrine 1,484 21% 2.1 Reportedly receiving their latrine from ide 1,019 14% 2.2 Reportedly obtaining a latrine directly from the LBO or an alternative source 463 6% 3. Potential wrongful cancellation to be investigated 186 3% TOTAL 7, % Confirmed cancellations Figure 13 presents the primary reason for cancellation as reported by customers with confirmed cancellation orders. Lack of sufficient finances was the primary reason for canceled orders. Roughly 9% of customers with canceled orders said they would be interesting in purchasing a latrine in the future on an installment plan and were subsequently followed up with by Provincial Financial Officers. F IGURE 13: R EASON FOR CANCELLATION Product/Service, 13% Other reasons, 2% Challenging environment, 10% Household residence issue, 14% Financial issue, 60%

19 Cambodia SMSU 2.0 Mid-term Results Page 19 Challenging Environment Issue: Some households live in areas with flooding, high ground water, bedrock or boulders underneath the soil preventing digging, or on a riverbank without sufficient land for installation. Financial Issue: Households in this category say their low finances is the reason for cancellation. Household Residence Issue: Households may rent their home or the land, which makes it unlikely for them to invest in a permanent latrine structure. Also, households in this category may be seasonally migrating outside of Cambodia or are currently building their house and are unwilling to purchase a latrine until the housing construction is complete. Product/ Service Issue: Households in this category were dissatisfied with the product design (i.e. 80cm ring) or delivery service (i.e. delays of delivery from LBO). Other Reasons: Most of the households in this category said they planned to use their neighbors or family member s latrine. Acquired latrine 21% of customers (1,484 total) reported receiving a latrine by the time the call center contacted them. After a modification to the call script, the call center determined that approximately 69% of this customer group eventually purchased a latrine through ide. The remaining 31% of customers purchased by directly contacting an ide LBO or an alternate source. Notably, 6% of the customers obtaining a latrine from a non-ide source reported doing so from a competing subsidized program for a lower price. This subsidized group makes up less than 1% of all customers with canceled orders. Further, we find that customers cancelling their latrine orders and eventually re-ordering their latrine from an ideaffiliated LBO, create duplicate orders that overestimate the cancellation rate. These customers created a 15% overestimation of cancellation during the WASH Call Center period, January 2016 to July As a result, this translates into an approximate 5% overestimation of the SMSU s overall cancellation rate. Wrongful cancellations Approximately 2.6% of contacted customers said they did not cancel their latrine order. Through further investigation, the team uncovered that out of all cancellations, approximately 1.4% were wrongful cancellations. These cases of wrongful cancellation are extremely rare. Additionally, their distribution in time and location do not support a hypothesis of malicious scheming. The systematic field investigation of the 186 cases found that there were logical, circumstantial reasons for the latrine cancellations. (For example, in some cases, the LBO gave up after multiple delivery postponements by the customer or there were genuine mistakes during backlog reconciliation.) Only 39 cases out of the 7,227 investigated were based on a unilateral LBO decision to avoid single deliveries in remote locations. These 39 cases are being dealt with by ide staff with care and respect for the households and further coaching for the concerned LBOs. Summary findings SMSU Management combined the WASH Call Center findings with the results from the independent field investigation conducted by SMSU M&E team. We conclude that: Both data sources identified that a substantial number of orders (an estimated 1,063 orders) listed as cancelled in Salesforce have actually been delivered by an ide-affiliated LBO. Through the field investigation into the 186 Suspicious Cancellation Cases, we confirm are approximately 88 cases of wrongful cancellation. Thus, the customers with canceled orders have been reclassified in Table 9 below. In summation, while investigating 7,227 cases of cancellation, the WASH Call Center identified that the minority of cancellations stemmed from customer dissatisfaction about long delivery times and product/service quality. The majority of cancellations were due to financial reasons. These findings invite SMSU management to continuously seek to improve SMSU s delivery mechanism, including adjusting the sanitation teacher commission-based system which has not changed since May 2013 and no longer reflects the 2017 market realities. However, concern about

20 Cambodia SMSU 2.0 Mid-term Results Page 20 possibility of widespread malicious schemes or fraudulent practices causing this cancellation rate can be put to rest based on the thorough analysis. T ABLE 9: R E-CLASSIFICATION OF CALL CENTER GROUPS 1. Confirmed Cancellations 5, % 2. Customer reporting having acquired a latrine from a non-ide affiliated source % 3. Wrongful order cancellations % Total 6, % DMQ RESULTS To better understand why households without latrines declined to order an SMSU latrine, the team launched a survey known as the Dark Matter Questionnaire or DMQ. After each sales pitch that results in a non-order, sanitation teachers now administer the DMQ survey. As of October 2017, the SMSU team has collected over 106,000 responses. The findings are presented below. Primary objection When asked about the primary objection to purchasing a latrine, 61% of all respondents reported that a financial issue was the main barrier, as shown in Figure 14. F IGURE 14: PRIMARY OBJECTION TO LATRINE PURCHASE, BY PROVINCE Banteay Meanchey 58% 21% 6% 7% 4% 3% Kampong Thom 54% 26% 7% 7% 2% 4% Kandal 48% 12% 11% 9% 13% 5% Oddar Meanchey 78% 6% 4% 7% 3% 1% Prey Veng 70% 9% 6% 4% 7% 4% Siemreap 57% 24% 7% 6% 3% 4% Svay Rieng 61% 16% 8% 5% 4% 6% ALL PROVINCES 61% 17% 7% 6% 5% 4% Financial Issue Other reasons Will install by myself Residence Issue Challenging environment Shared latrine Of the 106,000 plus DMQ records, we find that 16% of no sale interactions are with IDPoor1 households, 17% are with IDPoor2 households, and 67% are with non-poor households. Figure 15 shows us that while 71% of IDPoor1 households report financial issues as the primary objection, only 57% of non-poor households do. We would expect higher poverty households to have more financial limitations than lower, though interestingly even for nonpoor households, financial issues are still the main barrier.

21 Cambodia SMSU 2.0 Mid-term Results Page 21 F IGURE 15: PRIMARY OBJECTION TO LATRINE PURCHASE, BY IDPOOR 3% 4% 4% 4% 5% 5% 5% 5% 3% 7% 5% 9% 14% 16% 18% Shared latrine Challenging environment Residence Issue 71% 65% 57% Will install by myself Other reasons Financial Issue ID Poor 1 ID Poor 2 Non Poor Financial barrier To better understand the financial issues facing potential clients, our sanitation teachers then asked for the primary financial barrier. Of the subset of respondents that first reported financial issues, 51% said low income was the main barrier, as shown in Figure 16. Indebtedness was the second most common financial barrier, followed by the fact that the primary decision maker for large financial decisions was not present. F IGURE 16: PRIMARY FINANCIAL BARRIER, BY PROVINCE Banteay Meanchey 61% 10% 12% 5% 11% 1% Kampong Thom 47% 27% 16% 3% 4% 2% Kandal 68% 11% 15% 1% 3% 1% Oddar Meanchey 33% 30% 5% 30% 1% Prey Veng 52% 29% 11% 4% 1% 1% 2% Siemreap 47% 22% 15% 4% 11% 1% Svay Rieng 58% 12% 20% 1% 2% 4% 2% TOTAL 51% 22% 13% 7% 5% 1% 1% Very low income Indebtedness Decision maker not present Temporary cash flow issues Other financial Issue Saving for latrine & shelter Waiting on subsidy

22 Cambodia SMSU 2.0 Mid-term Results Page 22 We see in Figure 17 that the highest rates of low income barriers are among poor households. Interestingly, indebtedness does not seem to vary based on poverty of household. F IGURE 17: PRIMARY FINANCIAL BARRIER, BY IDPOOR 3% 3% 3% 1% 2% 6% 4% 5% 8% 7% 12% 22% 59% 25% 53% 15% 22% 48% Waiting on subsidy Saving for latrine & shelter Other financial Issue Temporary cash flow issues Decision maker not present Indebtedness Very low income Other barriers ID Poor 1 ID Poor 2 Non Poor While both residential issues and challenging environments rank low in terms of objection to purchase (6% and 5%, respectively), it is helpful to understand the driving factors behind those objections. For residential issues, we see in Figure 18 that lack of ownership or permanence are main limiting factors. Notably, we find that for 54% of IDPoor1 households, not owning land is the primary barrier, while that figure is only 32% for non-poor households. Further, for 42% of IDPoor2 and 43% of non-poor households, relocation of the household is the primary obstacle. F IGURE 18: PRIMARY RESIDENTIAL ISSUE BARRIER Household permanent relocation 42% Work migration House in construction 22% Not owning land 32% For challenging environments, we find that land size is the primary barrier faced by households, as shown in

23 Cambodia SMSU 2.0 Mid-term Results Page 23 Figure 19. These trends hold consistent across poverty groups. For this barrier it is interesting to note that while only 1% to 6% of households in all other provinces report issues with flooding, 22% of households in Oddar Meanchey list flooding as the primary challenging environment obstacle. This supports field findings that Oddar Meanchey is difficult province to sell in, primarily due to natural flooding. F IGURE 19: PRIMARY CHALLENGING ENVIRONMENT BARRIER Flood 3% Hard rock 9% High ground water 2% Small Land 58% House over riverbed 15% Household near water reservoir 11% BUSINESS ANALYSIS Summary of SMSU latrine business operators (LBOs) Sales records and coverage rates alone do not tell the whole story of the SMSU project. The program is also interested in the success of our micro-entrepreneurs, our LBOs. To better understand the strength of our LBOs, we conduct a business analysis of their operations. The data presented in Table 10 and Figure 20 is inclusive of LBO records from November 2014 to December In Table 10 we see that the average cost of a latrine unit is $52 while average selling price is $61, resulting in a $9 profit per unit. This averages out to a profit ratio of $0.14 for every dollar of sale. We see differences among the provinces in terms of average cost and profit. Siem Reap and Oddar Meanchey have the highest profit margins, due to both their low unit cost of production and higher price point. While Banteay Meanchey has the highest price point, it also has the highest production cost, thus reducing the overall profit margin. T ABLE 10: S NAPSHOT OF SMSU 2.0 LBOS INDICATORS VALUE NARRATIVE Total number of LBOs at the start of SMSU Total number of LBOs by the end of Total number of new LBOs recruited 58 Total number of sanitation teachers 327 This includes 172 LBOs working with the project and 46 LBOs discontinued/divested as of Nov 1, This includes 106 LBOs working with the project and 109 LBOs discontinued/divested as of Dec 31, This includes all new LBOs from Nov 1, 2014 Dec 31, These are sales agents involved with the LBOs by making sales as they receive a commission for every successful sale

24 Cambodia SMSU 2.0 Mid-term Results Page 24 Number of Active LBOs 111 Number of Inactive LBOs 104 Active LBOs are those who have made a sale in the past 6 months (as of Dec 2016), while the opposite is true for inactive LBOs. Currently, 52% of the LBOs are active while 48% are inactive Average cost of production of a latrine unit (All Provinces) $52 Banteay Meanchey $64 Kampong Thom $54 Kandal $49 Oddar Meanchey $59 Prey Veng $48 Siem Reap $50 Svay Rieng $48 Average selling price of a latrine unit (All Provinces) $61 Banteay Meanchey $71 Kampong Thom $63 Kandal $56 Oddar Meanchey $70 Prey Veng $55 Siem Reap $68 Svay Rieng $55 General estimation of operating profit ratio of a typical LBO $0.14 (All Provinces) Banteay Meanchey $0.09 Kampong Thom $0.13 Kandal $0.12 Oddar Meanchey $0.16 Prey Veng $0.13 Siem Reap $0.25 Svay Rieng $0.12 This includes labour, materials, commission and transport cost Average cost of production of a latrine unit by province. This is derived by averaging the unit sale price of the seven provinces Average selling price of a latrine unit by province. Roughly, an LBO makes 14 cents of profit for every dollar of sale Average profit for every dollar of sale by province. Note: In Banteay Meanchey LBOs have bigger slab (90cm square, 1m diameter ring, and square box) if compared to the rest of provinces. Average monthly volume of sale of Highest tier performers (Quintile 5) latrines sold on an average every month by the top 20% LBOs

25 Cambodia SMSU 2.0 Mid-term Results Page 25 Mid tier performers (Quintile 3) 16 Lowest tier performers (Quintile 1) 2 Average monthly revenue of Highest tier performers (Quintile 5) $4,088 Mid tier performers (Quintile 3) $1,001 Lowest tier performers (Quintile 1) $113 Average monthly net profit of Highest tier performers (Quintile 5) $648 Mid tier performers (Quintile 3) $138 Lowest tier performers (Quintile 1) $15 17 latrines sold on an average every month by the middle 20% LBOs 2 latrines sold on an average every month by the bottom 20% LBOs Net sales or revenue (units sold times unit sale price) earned every month by the three groups Average net profit (gross revenue minus cost of goods sold) earned every month by the three groups Figure 20 presents the average sale price for a latrine unit, broken down by material, labor, transport, and commission costs to the LBO, as well as their profit margin. Materials and labor are primary costs faced by LBOs. It is interesting to note that Oddar Meanchey has the highest transportation costs, which is to be expected given our understanding of the challenging environment in that province. $80 F IGURE 20: PRICE COMPOSITION OF LATRINE SET FROM LBO (INCLUDES PROFIT) $70 $60 $50 $40 $30 $6.75 $5.00 $4.38 $23.88 $8.02 $5.00 $4.00 $19.00 $6.87 $5.00 $2.50 $16.50 $11.13 $5.00 $6.00 $19.25 $6.98 $5.00 $3.00 $16.25 $17.07 $6.78 $5.00 $5.00 $3.25 $3.00 $17.38 $16.25 $20 $10 $31.25 $26.48 $25.38 $28.62 $23.78 $24.80 $23.97 $0 Banteay Meanchey Kampong Thom Kandal Oddar Meanchey Prey Veng Siemreap Svay Rieng Material Labor Transport Commission Margin Gender analysis In addition to standard LBO performance such as profitability, we wanted to better understand the role that women play in the LBO business. Based on field reports, we learned that while few women were not the primary LBO owner (9% of all LBOs are female), many play a role in the LBO businesses. ide hypothesized that LBO businesses with active female involvement were better administered than those without. To test this theory, SMSU coordinating staff (District Supply Chain Coordinators - DSCCs) completed two assessments from June 12 to 20, They first rated the degree of female involvement in the LBO business on an ordinal scale from one to six. DSCCs were then asked to rate the LBO s business performance on an ordinal

26 Cambodia SMSU 2.0 Mid-term Results Page 26 scale from one to four. Factors considered when measuring LBO performance include quality of their record keeping and client support, inventory management, working capital and cash management, reliability, and willingness to take risk and/or change. The DSCC score of LBO performance was then combined with sales data such as cancellation rates, delivery rates, and order-to-delivery time. Our sample of eligible LBOs for analysis resulted in a sample size of only 60 LBOs. Preliminary correlation and regression analysis did not find statistically significant results, though the direction of the relationship is what we hypothesized. That is as level of female involvement goes up, so does the LBO s performance score. Recommendations for future analysis include a more nuanced measure of female involvement, as well as expanding the sample size to increase statistical rigor. FINAL NOTE These analyses are not exhaustive of all M&E activities in SMSU 2.0. Instead, they are meant to provide a helpful mid-term update to the project for monitoring and evaluation. While these findings shed light on many important components of SMSU, they also serve to guide future M&E activities as new questions emerge. Further, insights learned from this report will be used to adjust program activities as needed for final project implementation.