Exploratory Data Analysis of WSS Operators Sample in Brazil

The observation of survey outcomes will compare across potential relevant features:

  • Ownership/type and
  • Size of covered area
  • Service mix (W/S/W+S)
  • (possibly) Prevalence of urban vs rural coverage vs
  • (Age)
  • (State)

Here is a very basic distribution of our sampled Suppliers in Brazil, by STATE (UF), Ownership Type and size of Coverage reach

load(here::here("output", "tbl_uf_own_cov.Rdata"))
# class(tbl_uf_own_cov)
caption <- "Sampled Suppliers in Brazil, by STATE (UF), Ownership Type and size of Coverage reach"
covtable <- pandoc.table.return(tbl_uf_own_cov, 
                                          keep.line.breaks = F, 
                                          style = "multiline", 
                                          justify = "lcrr", 
                                          caption = caption
                                          )

cat(covtable)
Sampled Suppliers in Brazil, by STATE (UF), Ownership Type and size of Coverage reach
NM_ESTADO Q1_ownership_lbl Cov_pop_rango_lbl n
ACRE Public Large 500k-1m 1
ALAGOAS Other Large 1m-10m 1
AMAPÁ Other Large 500k-1m 1
AMAZONAS Public Medium 50-100k 1
AMAZONAS Other Medium 300-500k 1
BAHIA Other Mega >10m 1
CEARÁ Other Large 1m-10m 1
DISTRITO FEDERAL Other Large 1m-10m 1
ESPÍRITO SANTO Other Large 1m-10m 1
GOIÁS Public Small 10-50k 1
MARANHÃO Other Large 1m-10m 1
MATO GROSSO Public Very Small <10k 1
MATO GROSSO Public Medium 200-300k 1
MINAS GERAIS Public Small 10-50k 1
MINAS GERAIS Other Mega >10m 1
PARÁ Public Large 1m-10m 1
PARAÍBA Other Large 1m-10m 1
PARANÁ Other Mega >10m 1
PERNAMBUCO Other Large 1m-10m 1
RIO DE JANEIRO Public Medium 300-500k 1
RIO DE JANEIRO Other Mega >10m 1
RIO GRANDE DO NORTE Other Large 1m-10m 1
RIO GRANDE DO SUL Public Medium 50-100k 1
RIO GRANDE DO SUL Public Large 1m-10m 1
RIO GRANDE DO SUL Private Medium 50-100k 1
RIO GRANDE DO SUL Other Large 1m-10m 1
RONDÔNIA Public Large 1m-10m 1
SANTA CATARINA Public Small 10-50k 1
SANTA CATARINA Public Medium 50-100k 1
SANTA CATARINA Public Large 500k-1m 1
SANTA CATARINA Private Large 1m-10m 1
SÃO PAULO Public Very Small <10k 1
SÃO PAULO Public Small 10-50k 1
SÃO PAULO Public Medium 200-300k 1
SÃO PAULO Public Mega >10m 1
SÃO PAULO Private Large 1m-10m 2
SÃO PAULO Other Mega >10m 1
SERGIPE Other Large 1m-10m 1
Total - - 39

Legislative Context

  1. Overall, the majority of the Suppliers declare they operate in a context where there is legislation for key operational aspects such as:
    • Coverage Areas
    • Service
    • Quality
    • Tariffs
Plot_CountByLawOwnersh.png
Plot_CountByLawOwnersh.png

1) COVERAGE & SERVICE

Summary of key coverage numeric variable

load(here::here("output", "tbl_descstat_cov.Rdata"))
# class(tbl_descstat_cov)

caption <- "Summary of Coverage variables"
var.summary <- pandoc.table.return(tbl_descstat_cov , 
                                              keep.line.breaks = TRUE,
                                              round = 2,
                                              big.mark = ",",
                                              justify = "llrrrrrr", # OKKIO AL # DI COLONNE
                                              caption = caption, style = "multiline",
                                              split.table = Inf
                                              )

cat(var.summary)
Summary of Coverage variables + The WSS suppliers in the sample - as of April 21, 2022 - cover an average of 88.4615385 municipios (almost always in one State), corresponding to a mean covered population of 4.53022^{6}. + The suppliers on average serve localities that are 83.9184211 urban.
Variable unit_of_analysis N Mean Std. Dev. Median Min Max
N_cov_states NA 39 1.79 2.47 1 1 12
N_cov_mun Num of Municipio covered (Annex) 39 88.44 140.6 22 1 629
Q2_CovMun_decl Num of Municipio covered (Respondent) 39 88.46 140.6 22 1 629
totpop2018 Tot Pop in covered Mun (IBGE2018) 39 4,530,220 6,759,477 1,707,440 3,928 29,879,718
Q44_Past12m_Interr_RespTime NA 31 1.79 3.41 1 0.04 18.91
Q5a_Urb_Perc2 NA 19 83.92 14.87 90 52 99

Area(s) covered vs. service

Is there any difference between the Product MIX Q9a_d_Mix_comb vs the URB-RUR MIX of the supplier Q4_UrbRur_Mix1?

There are no suppliers serving exclusively rural area. No remarkable differences in the Product MIX across the URB / RUR MIX types

Plot_CountByProdMix.png

NEXT:… (?) Is the urbanization process (size, local distribution, rate) causing the lack (access, quality, affordability ) to service in the region (controlling for physical assets / hydrological condition, pop #, etc) ? ??

Summary of key service charateristics

the WSS suppliers in the sample as of April 21, 2022, …

load(here::here("output", "tbl_descstat_covserv.Rdata"))
# class(tbl_descstat_cov)

caption <- "Summary of Service variables"
var.summary <- pandoc.table.return(tbl_descstat_covserv , 
                                              keep.line.breaks = TRUE,
                                              round = 2,
                                              big.mark = ",",
                                              justify = "llrrrrrr", # OKKIO AL # DI COLONNE
                                              caption = caption, style = "multiline",
                                              split.table = Inf
                                              )

cat(var.summary)
Summary of Service variables
Variable unit_of_analysis N Mean Std. Dev. Median Min Max
Q7a_Volume_Res_Perc % of m3 38 87.27 7.22 89 60 95.92
Q7b_Volume_Com_Perc % of m3 37 7.91 4.26 7.5 0.13 20
Q7c_Volume_Ind_Perc % of m3 37 1.89 2.9 0.7 0 13.3
Q7d_Volume_Inst_Perc % of m3 37 3.14 2.89 2.05 0 13.3
Q10a_Wconn_Res Num connections water 38 799,607 1,388,985 215,680 1,040 7,172,000
Q10b_Wconn_Tot Num connections water 39 1,008,353 1,738,252 314,182 1,485 9,053,000
Q10c_Wconn_Act Num connections water 39 888,812 1,549,006 295,460 1,190 7,987,000
Q11a_Sconn_Res Num connections sanitation 32 444,860 1,104,293 79,099 2,097 5,984,000
Q11b_Sconn_Tot Num connections sanitation 35 597,450 1,362,863 106,000 2,275 7,495,000
Q11c_Sconn_Act Num connections sanitation 34 568,008 1,256,892 118,875 2,198 6,718,000
Q12a_m3_In_Res m3 water entering system 20 127,187,676 213,537,966 38,813,618 467,500 833,099,200
Q12b_m3_In_Tot m3 water entering system 39 210,130,900 3.37e+08 61,092,028 190,600 1.631e+09
Q13_m3_Fatt_Tot m3 water billed 39 1.18e+08 182,513,737 24,591,156 123,600 785,720,000
Q13a_m3_Fatt_Res_Perc % of m3 water billed 38 85.31 9.31 87.33 52 95.23
Q13b_ReaisFatt_Tot Revenue in Reais 38 746,668,127 1.177e+09 214,512,932 546,573 4.601e+09
Q13c_ReaisFatt_Res_Perc % Residential of revenue in Reais 38 79.07 9.4 78.38 56.46 95
NRWm3_pct % of m3 Water (in the system) NOT billed 39 38.84 16.13 37.38 6.38 69.66
Q23_Network_km km 39 80,829 308,791 4,158 9 1,828,660
Q27_IndivMeter_Perc % of connections 38 86.3 24.1 99 12 100
Q40_NetworkYrlyInspected_Perc % of network 22 70.73 37.13 89.5 2.3 100
Q43_Past12m_Interr_PercAff % 25 38.51 34.48 30 0 100

NRW % levels

Lit Input: Some key determinants of high NRW are: 1. The cost of implementing counter-measures 2. A political cost associated with the control of unauthorized consumption in marginalized ares 3. A cost to those people who benefit from corruption 4. Technical cost of shortening leakage detection time & improving pressure management (especially when the cost of pumping more raw water is 0 or low) [see González-Gómez, García-Rubio, and Guardiola (2011), pp. 33-35; ch. 1].

Based on the information collected via the survey, we can estimate NRW as per International Water Association’s (IWA) definition: NRW is defined as water that is placed into a water distribution system that is not billed to customers. Our derived indicator is NRWm3_pct { = (Q12b_m3_In_Tot - Q13_m3_Fatt_Tot) / Q12b_m3_In_Tot *100 }

load(here::here("output", "BRA_NRW.Rdata"))
 
# class(tbl_descstat_cov)

# SIMPLE TABLE WITH FILTER 
BRA_NRW %>% dplyr::select(-M1_oper_id, -Q12b_m3_In_Tot, -Q13_m3_Fatt_Tot) %>% 
    dplyr::arrange(NM_ESTADO) %>% 
    DT::datatable(., 
                  elementId = NULL, 
                 # filter = 'top',  # Problem if there are empty cells !!!
                  extensions = 'Buttons', 
                  options = list( pageLength = 50,searchHighlight = TRUE,   dom = 'Bfrtip', buttons = c( 'csv', 'excel','pdf'))
    ) %>%
    # helper functions 
        DT::formatRound('NRWm3_pct', 1) %>% 
        DT::formatStyle( 'NRWm3_pct', backgroundColor =  'yellow') %>% 
        DT::formatStyle( 'NRWm3_pct', color = styleInterval( 50, c('black', 'red')) )

Initial exploration shows that:

  1. On average, the % NWR varies quite dramatically across different ownership types (PUBLIC-owned 32% < Other 38% < Private 60% (more dispersed))
  2. Mean % NWR is lower for suppliers exclusively serving urban areas (34%) than for suppliers servin mixed-urb-rur (43%) … but sems quite disperserd.
  3. On average, the size of served population, is also a great source of variation in the %NRW - highest found in 500-900k size (58%) and smallest in 10-50k size (7%)
Plot_DensNRW_byOwn.png
Plot_DensNRW_byLOC.png
Plot_DensNRW_bySIZE.png

NRW correlation with “suspected” determinant factors

Visually explore whether the % of NRW shows any type of correlation with some variables that could be relevant:

  1. Q27_IndivMeter_Perc –> no stryking connection
Plot_grid_NWR_IndMetBY4.png
  1. Q40_NetworkYrlyInspected_Perc –> no stryking connection
Plot_grid_NWR_NetInpsBY4.png
  1. Q24_VolLossMech_Has –> most suppliers say “Yes” (have a system to record VOLUME LOSSES)
Plot_BOXNWR_VolLoss.png
  1. Q39_LeaksDetection_Analyzed –> most suppliers say “Yes” (have a systematic campaign to detect ruptures and LEAKS) and those cases show a somewhat lower NRW
Plot_BOXNWR_Leaks.png
  1. Q41_ClandConn_Analyzed –> the vast majority suppliers say “Yes” (have a system) and those cases show a significantly lower NRW (for now the NO sub-sample is too small to further analyze)
Plot_BOXNWR_Cland.png

NEXT … Further exploration ….. + Exist procedim for estimation of vol loss/ (Q24)
* (?) Is the year of operation anywhere connected with efficiency?
* (?) Is thestack of laws connected with efficiency (Expected inverse relation)?? Q8a_c_Leg_comb Q8a_LegInst_Cover Q8b_LegInst_DirServ “Q8c_LegInst_DirQual Q8d_LegInst_Tarif,”

Reliability of the service (Interruptions)

Lit Input: Performance can improve with: 1) carefully designed private sector participation in service provision 2) regulatory agencies that are transparent, accountable, and free of political interference, 3) strong accountability mechanisms for SOE. (see Andrés, Schwartz, and Guasch 2013 ).

Y + Reliability = Q4… Q45 interruptions X + location + size + age? Q6 + Quality regulation (q8c) + SWIT penetration ? + % eng in staff + % IT in staff?

The table below shows responses related to the incidence of service interruptions in the past 12 months.

load(here::here("output", "BRA_REL.Rdata"))
 
# class(tbl_descstat_cov)

# SIMPLE TABLE WITH FILTER 
#datatable(data= BRA_REL  , filter = 'top')
DT::datatable(BRA_REL, 
                  elementId = NULL, 
                  #filter = 'top',  
                  extensions = 'Buttons', 
                  options = list( pageLength = 50,  dom = 'Bfrtip', buttons = c( 'csv', 'excel')),
                  caption = 'Individual Supplier\'s Responses about Service Interruptions'
) %>%
    formatRound("Perc of network affected", 1) %>%
    formatRound("Avg # days of interruption", 1) %>% 
        DT::formatStyle("Perc of network affected", color = styleInterval( 50, c('black', 'red')) )

The chart below shows the most frequent causes for service interruption p

InterruptReasons

2) SWIT

Use of SWIT still not widespread for various barriers: + high capital investments + slow cost recovery + limited technical capacity, particularly for small utilities, + lack of integration between different technologies

Summary of SWIT adoption in the Brazil sample

The table belows shows to what degree the WSS suppliers in the sample nrow(BRA_REL) already adopt various types of Smart Water Management: dedicated R&D and IT departments, Network Management systems, GIS, Hidraulic Models, DMA & pressure management, and Smart meters.

The % of adoption (albeit Count includes when tools are still “in construction”) appear quite high.

NOTES + Some of the respondents are actually large companies comglomerate, so will have to understand better their distribution by size and ownership…. + Even when companies say “yes” to the adoption (e.g. smart meter) it may be just a pilot…

TblSWIT_tbl.png

Below is a different graph to show the degree of adoption of various types of technologies.

ggSWITbar.png

3) CORPORATE GOVERNANCE

We learned from the survey who is entitled to hire & fire the CEO / President of the agency, his/her remuneration the Sr. Directors.

Appointment of CEO/President/Sr Directors in the Brazil sample

The plots below compare the above across types of ownership and size of the covered area.

  • It would appear that the “Prefeito/Gobernador” is in charge in most of the public agencies
  • In privately owned suppliers, the Board seems the have the responsability monst cases
Q46hireCEO_own_plot.png
Q49fireCEO_own_plot.png
Q46hireCEO_size_plot.png
Q49fireCEO_size_plot.png
Q48Dir_own_plot.png
Q48Dir_size_plot.png

4) FINANCIAL HIGHLIGHTS

Lit Input: 1) under-investment + dilapidated infrastructure –> INEFFICIENCIES (see Arniella 2018 ) 2) Energy consumption much of O&M costs (see Lackey and Fillmore 2017)

OPEX by source of financing

Below are the findings from the Brazil sample by group.

load(here::here("output", "tbl_OPEX.Rdata"))
# class(tbl_descstat_cov)

caption <- "Average composition of OPEX - grouped by operator type"
 
tbl_OPEX_site <-    pandoc.table.return(tbl_OPEX,
                              keep.line.breaks = TRUE,
                              split.tables = Inf,
                              round = 2, big.mark = ",",
                              justify = "lllrrrr", 
                              caption = caption, 
                              style = "simple",
                              missing = "")  %>% 
    cat()
Average composition of OPEX - grouped by operator type
Ownership condition Population coverage size % OPEX from tariffs (mean) % OPEX from subsidy (mean) % OPEX from lending (mean) % OPEX from other (mean) N
Public Small0_50k 100 0 0 0 6
Medium50_500k 98.33 0.83 0.83 0 6
Large500_10m 100 0 0 0 5
Mega_up10m 1
Total 99.33 0.33 0.33 0 18
Private Medium50_500k 100 0 0 0 1
Large500_10m 90 0 10 0 3
Total 92.5 0 7.5 0 4
Other Medium50_500k 100 0 0 0 1
Large500_10m 95.78 4.22 0 0 11
Mega_up10m 100 0 0 0 5
Total 97.27 2.73 0 0 17
Total 97.6 1.43 0.97 0 39

CAPEX by source of financing

Below are the findings from the Brazil sample by group.

load(here::here("output", "tbl_CAPEX.Rdata"))
# class(tbl_descstat_cov)

caption <- "Average composition of CAPEX - grouped by operator type"

tbl_CAPEX_site <- 
    pandoc.table.return(tbl_CAPEX,
    keep.line.breaks = TRUE,
    split.tables = Inf,
    round = 2, big.mark = ",",
    justify = "lllrrrr",
    caption = caption,
    style = "simple",
    missing = "") %>%
  cat()
Average composition of CAPEX - grouped by operator type
Ownership condition Population coverage size % CAPEX from tariffs (mean) % CAPEX from subsidy (mean) % CAPEX from lending (mean) % CAPEX from other (mean) N
Public Small0_50k 100 0 0 0 6
Medium50_500k 92 7 1 0 6
Large500_10m 50.83 37.87 11.3 0 5
Mega_up10m 0 0 100 0 1
Total 73.86 13.51 12.63 0 18
Private Medium50_500k 50 50 0 0 1
Large500_10m 40 0 60 0 3
Total 43.33 16.67 40 0 4
Other Medium50_500k 100 0 0 0 1
Large500_10m 34.54 40.85 10.34 15.3 11
Mega_up10m 34.5 10 43 12.5 5
Total 39.57 32.96 14.92 13.69 17
Total 53.96 23.23 16.84 6.59 39

Foreign participation in capital

Only 6 (out of 39) companies declare they have participation of foreign capital and - of those - the average share is 35%.

5) TARIFFS

IADB (?): we have no idea if tariffs are low or high in LAC (people say are too low, but in fact they are not showing in the Database)

Fixed tariff component

load(here::here("output", "tbl_Fixed_Crit.Rdata"))
# class(tbl_descstat_cov)
caption <- "Determinants of Tariff Fixed Components"
tbl_Fixed_Crit_site <-  pandoc.table.return(tbl_Fixed_Crit,
                                                 keep.line.breaks = TRUE,
                                                 split.tables = Inf,
                                                 round = 2, big.mark = ",",
                                                 justify = "lcrrrr", 
                                                 caption = caption, 
                                                 style = "simple"#,
                                                 #missing = ""
                                                 )  %>%     cat()
Determinants of Tariff Fixed Components
Q14_Tar_Fix_Has Fixed Component? (N) Flat Income/Social Location type Consumer type
Yes 34 14.71% 79.41% 17.65% 88.24%
No 5 NA% NA% NA% NA%

Variable tariff component

load(here::here("output", "tbl_Variable_Crit.Rdata"))
# class(tbl_descstat_cov)
caption <- "Determinants of Tariff Variable Components"
tbl_Variable_Crit_site <-   pandoc.table.return(tbl_Variable_Crit,
                                                         keep.line.breaks = TRUE,
                                                         split.tables = Inf,
                                                         round = 2, big.mark = ",",
                                                         justify = "lcrrrrrr", 
                                                         caption = caption, 
                                                         style = "simple"#,
                                                         #missing = ""
)  %>%  cat()
Determinants of Tariff Variable Components
Q15_Tar_Var_Has Variable Component? (N) Constant Marginal Increasing Blocks Decreasing Blocks Income/Social Consumer Location Consumer type
Yes 39 2.56% 94.87% 5.13% 92.31% 7.69% 94.87%

Latest 3 changes in tariffs

The survey asked the WSS providers the date and % change of the three more recent tariff changes. The plots below show the distribution of the latest adjustment (percentage change to average tariff) across types of ownership and size of the covered area. Depending on the operator:

  • “t-2” was some time between 2008 and 2017
  • “t-1” was some time between 2007 and 2018
  • “t” (or the latest change) was some time between 2014 and 2019
TarChange_p.png
TarChange_p2.png

REFERENCE

Andrés, Luis A., Jordan Schwartz, and J. Luis Guasch. 2013. Uncovering the Drivers of Utility Performance: Lessons from Latin America and the Caribbean on the Role of the Private Sector, Regulation, and Governance in the Power, Water, and Telecommunication Sectors. The World Bank. https://doi.org/10.1596/978-0-8213-9660-5.
Arniella, Elio. 2018. “CAN SMART TECHNOLOGIES IMPROVE EFFICIENCY IN THE WATER AND SANITATION SECTOR? Pressure Management Case Studies.” Nota {{Tecnica}}. IADB.
González-Gómez, Francisco, Miguel A. García-Rubio, and Jorge Guardiola. 2011. “Why Is Non-Revenue Water So High in So Many Cities?” International Journal of Water Resources Development 27 (2): 345–60. https://doi.org/10.1080/07900627.2010.548317.
Lackey, Katy, and Lauren Fillmore. 2017. “Energy Management for Water Utilities in Latin America and the Caribbean - Exploring Energy Efficiency and Energy Recovery Potential in Wastewater Treatment Plants.” Water Environment & Reuse Foundation (WE&RF) for World Bank Group.

  1. Later on, I would like to explore if (?) Is the urbanization process (size, local distribution, rate) causing the lack (access, quality, affordability ) to service in the region (controlling for physical assets / hydrological condition, pop #, etc) ? ??↩︎

Work in progress, not (yet) ready for dissemination.