Valuing Water Purification by Forests: An Analysis of Malaysian Panel Data


Water purification might be the most frequently invoked example of an economically valuable ecosystem service, yet the impacts of upstream land use on downstream municipal water treatment costs remain poorly understood. This is especially true in developing countries, where rates of deforestation are highest and cost-effective expansion of safe water supplies is needed the most. We present the first econometric study to estimate directly the effect of tropical forests on water treatment cost. We exploit a rich panel dataset from Malaysia, which enables us to control for a wide range of potentially confounding factors. We find significant, robust evidence that protecting both virgin and logged forests against conversion to nonforest land uses reduced water treatment costs, with protection of virgin forests reducing costs more. The marginal value of this water purification service varied greatly across treatment plants, thus implying that the service offered a stronger rationale for forest protection in some locations than others. On average, the service value was large relative to treatment plants’ expenditures on priced inputs, but it was very small compared to producer surpluses for competing land uses. For various reasons, however, the latter comparison exaggerates the shortfall between the benefits and the costs of enhancing water purification by protecting forests. Moreover, forest protection decisions that appear to be economically unjustified when only water purification is considered might be justified when a broader range of services is taken into account.

This is a preview of subscription content, log in to check access.

Fig. 1
Fig. 2
Fig. 3
Fig. 4


  1. 1.

    The process-based model InVEST is increasingly being used to predict water treatment costs under different land-use scenarios (Conte et al. 2011).

  2. 2.

    Treatment plants in Perak measure the concentration of suspended solids in source water as it enters the plant, and they adjust the amount of alum, the chief chemical used to remove suspended solids, accordingly. Alum reduces the pH of treated water, thus requiring the addition of lime before the water is distributed and causing a knock-on increase in treatment costs (Mohd Nordin et al. 2000). The third major treatment chemical is chlorine, a disinfectant, but treatment plants in Perak add it as a standard dose that does not vary with source water quality (personal communication, Perak Water Board, February 10, 2014).

  3. 3.

    The proportion of the population with access to piped water or another improved water source is higher in Malaysia than in most developing countries: 88.2 % in 1990 and 99.6 % in 2010, according to the World Bank’s World Development Indicators database ( Currently, 98 % of Perak households have a piped water connection (personal communication, Perak Water Board, May 26, 2015).

  4. 4.

    This double-log specification for treatment cost and water volume is common in the econometric literature on water treatment (Forster et al. 1987; Moore and McCarl 1987; Holmes 1988; Murray and Forster 2001; Piper 2003). We confirmed the appropriateness of the log-transform of treatment cost by estimating a Box-Cox model, which favored that specification over a linear specification: the estimated theta parameter in a left-hand-side Box-Cox model was 0.117, which is much closer to 0 than to 1. We confirmed the appropriateness of the log-transforms of the other explanatory variables by comparing the overall fit of models containing the transformed variables to ones containing untransformed variables.

  5. 5.

    Some prior studies have used unit cost (C / Q) as the dependent variable (e.g., Dearmont et al. 1998; Piper 2003; Ernst 2004). Our double-log specification encompasses this as a special case \((\alpha = 1)\).

  6. 6.

    In addition, data were incomplete for some TPs: 32 % of the TP-years had fewer than 12 months of data, with 19 % having fewer than 11 months. It is possible that some of the missing observations resulted from outages or failures caused by poor source water quality. We doubt that this explains most of the missing observations, however, which were typically missing for all or nearly all of the TPs in the state or a particular region of the state in a given month. This pattern implies problems with data recording and storage, not temporary closures of individual TPs.

  7. 7.

    This problem can be avoided by estimating a profit function instead of a cost function. See Pattanayak and Kramer (2001) for an application to a different watershed service, drought mitigation. We were unable to estimate a profit function because the Perak Water Board did not provide data on revenue from water sales.

  8. 8.

    A private company, the Malaysian Utilities Corporation, operates two additional TPs in Perak (Sultan Idris Shah II, Ulu Kinta). It operated a third until 2001 (Pekan Parit). A second company, GSL Water, operates two others (Gunung Semanggol, Taiping Headworks). We excluded these private TPs from the analysis due to their different management and because the companies did not report plant-level data in all time periods.

  9. 9.

    Air Kuning, Gerik V, Jalan Baru, and Pulau Banding. The exclusion of TPs that used reservoirs probably biased our estimate of the effect of forests on treatment costs downward, as we have ignored dredging costs.

  10. 10.

    Selama, which is one of the newest TPs (established in 2004).

  11. 11.

    Grik, Sg. Jelintoh, and Ulu Soh, which closed in 2000, and Felda Bersia, which closed in 2001.

  12. 12.

    BB Seri Iskandar (opened in 2001), Sg. Tapah (2002), Tg. Malim (2003), and Hilir Perak (2006).

  13. 13.

    Perlop (1995), Terong (1999), Padang Rengas (2001), Teluk Kepayang (2001), Kota Lama Kiri (2002), Sg. Kampar (2004), Kg. Paloh (2005), Pengkalan Hulu (2005), Jelai (2007), Sumpitan (2007), and Manong (2008).

  14. 14.

    A complete list is available at:

  15. 15.

    A complete list is available at:

  16. 16.

    Data on other variables, such as the components of total operating cost (labor, energy, chemicals, maintenance) or quantities of specific treatment chemicals, were too incomplete to analyze.

  17. 17.

    We dropped 13 observations that appeared to be affected by data-entry error, as indicated by water volumes or unit costs that were either more than twice as large or less than half as large as any other values for the same TPs. These observations came from 11 TPs. Dropping them had a negligible impact on the regression results.

  18. 18.

    The NFIs contain information on the approximate time periods when logged forests were harvested. The different NFIs do not define all the time periods consistently, however. Furthermore, for time periods defined consistently between NFI 3 and NFI 4, the reported areas exhibit substantial logical inconsistencies in many catchments (e.g., the area classified as having been logged before a given year is reported as being larger in NFI 4 than in NFI 3). For these reasons, we did not analyze the effect of years since logging on treatment cost. This effect is likely negative: sediment loads decrease relatively rapidly after logging in tropical forests, though they remain higher than sediment loads from virgin forests for decades (Bruijnzeel 2004; Abdul Rahim and Zulkifli 2004).

  19. 19.

    The sample for this calculation included only countries that reported a positive virgin forest area in 1990. The UN estimates refer to primary and secondary forests, which we assumed correspond to virgin and logged forests as defined in Malaysia. The endpoints, 1990 and 2005, are the UN reporting years closest to NFI 3 (1992) and NFI 4 (2004).

  20. 20.

    Holmes (1988) reported a much lower turbidity elasticity, \(-\)0.07.

  21. 21.

    Fixed effects were more appropriate than random effects because the sample included nearly all TPs in Perak, not a random selection of them (Wooldridge 2010, pp. 285–287; Kennedy 2008, p. 291). Moreover, Hausman tests rejected the null that the random effects were uncorrelated with the error term: \(\chi ^{2}(24) = 95.3\,(P = 0.000)\) if the test used the disturbance variance estimate from the random effects model and \(\chi ^{2}(24) = 97.0\,(P = 0.000)\) if it used the estimate from the fixed effects model. Despite these test results, the random effects coefficient estimates were not very different from the fixed effects estimates given by model (5) in Table 2: \(-\)1.41 \((P = 0.000)\) for virgin forest and \(-\)1.06 \((P = 0.000)\) for logged forest.

  22. 22.

    Endogeneity of a different sort—a correlation between the forest cover variables and the error term—could occur if the four TPs that were established after 1994 were designed to cope with the effects of the land-use changes that occurred during the sample period. Reestimating model (5) with those TPs excluded from the sample did not change the results much: the coefficients on virgin forest and logged forest were \(-\)1.56 (\(P = 0.019\)) and \(-\)1.13 (\(P = 0.0.34\)), respectively.

  23. 23.

    Results are very similar to those for model (6) if we use the first lag as an IV: the coefficients on virgin forest and logged forest are \(-\)1.62 \((P = 0.009)\) and \(-1.18\,(P = 0.020)\), respectively. The instrument is again very strong \((F = 1989)\).

  24. 24.

    This is also the case if we use the first lag as an instrument \((P = 0.711)\).

  25. 25.

    Air Ganda, BB Seri Iskandar, Kg. Gajah, Kg. Paloh, Kota Lama Kiri, and Teluk Kepayang.

  26. 26.

    Table 1 in FAO and CIFOR (2005) indicates that land use affects sediment loads only in catchments up to 10,000 ha, which is smaller than 40 % of the observations in our sample. Ogden et al. (2013) note that the table is drawn from a source that provides no citations to support this threshold.

  27. 27.

    An additional explanation could be that the US elasticities are biased upwards by a correlation between water volume and some omitted variable. A strong candidate for such a variable is catchment area. In our sample, water volume has a relatively large positive correlation with catchment area \((\rho = 0.310, \,P = 0.000)\). This correlation is not surprising; larger TPs require larger supplies of source water and thus have larger catchments. The US studies do not control for catchment area, but model (5) controls for it through the TP effects.

  28. 28.

    The sign pattern on the linear and quadratic terms is the opposite (positive and negative) if TP effects are excluded, as in model (3). This demonstrates the importance of controlling for unobserved TP characteristics.


  1. Abdul Rahim AS, Mohd Shahwahid HO (2011) A panel data analysis of timber harvesting operations and its impact on the cost of water treatment. Aust J Basic Appl Sci 5:598–601

    Google Scholar 

  2. Abdul Rahim N, Zulkifli Y (2004) Hydrological impacts of forestry and land-use activities: Malaysian regional experience. In: Abdul Rahim N (ed) Water: forestry and land use perspectives, Technical Documents in Hydrology No. 70, International Hydrological Programme. UNESCO, Paris, pp 86–105

    Google Scholar 

  3. Alcott E et al (eds) (2013) Natural and engineered solutions for drinking water supplies: lessons from the northeastern United States and directions for global watershed management. CRC Press, Boca Raton

  4. Angrist JD, Pischke JS (2009) Mostly harmless econometrics: an empiricist’s companion. Princeton University Press, Princeton

    Google Scholar 

  5. Barten PK, Ernst CE (2004) Land conservation and watershed management for source protection. J Am Water Works Assoc 96:121–135

    Google Scholar 

  6. Beck NL, Katz JN (1995) What to do (and not to do) with time-series cross-section data. Am Polit Sci Rev 89:634–647

    Article  Google Scholar 

  7. Bennett G et al (2013) Charting new waters. Forest Trends, Washington

    Google Scholar 

  8. Bruijnzeel LA (2004) Hydrological functions of tropical forests. Agric Ecosyst Environ 104:185–228

    Article  Google Scholar 

  9. Carlson KM et al (2014) Influence of watershed-climate interactions on stream temperature, sediment yield, and metabolism along a land use intensity gradient in Indonesian Borneo. J Geophys Res Biogeosci 119:1110–1128

    Article  Google Scholar 

  10. Chichilnisky G, Heal GM (1998) Economic returns from the biosphere. Nature 391:629–630

    Article  Google Scholar 

  11. Conte M et al (2011) Retention of nutrients and sediment by vegetation. In: Kareiva P et al (eds) Natural capital: theory and practice of mapping ecosystem services. Oxford University Press, Oxford, pp 89–110

    Google Scholar 

  12. Dearmont D et al (1998) Costs of water treatment due to diminished water quality: a case study in Texas. Water Resour Res 34:849–853

    Article  Google Scholar 

  13. Department of Statistics (2010) Input-output tables, Malaysia 2005: volume 2. Putrajaya, Malaysia

  14. Dissmeyer GE (ed) (2000) Drinking water from forests and grasslands: a synthesis of the scientific literature. Gen Tech Rep SRS-39, USDA Forest Service, Southern Research Station, Asheville, North Carolina

  15. Dudley N, Stolton S (2003) Running pure: the importance of forest protected areas to drinking water. World Bank and WWF, Washington

    Google Scholar 

  16. Dunne T (1979) Sediment yield and land use in tropical catchments. J Hydrol 42:281–300

    Article  Google Scholar 

  17. Dunne T, Leopold LB (1978) Water in environmental planning. WH Freeman, New York

    Google Scholar 

  18. Ernst C (2004) Protecting the source: land conservation and the future of America’s drinking water. Trust for Public Land, San Francisco

    Google Scholar 

  19. Ernst C et al (2004) Protecting the source: conserving forests to protect water. Opflow 30(1):4–7

    Google Scholar 

  20. FAO (2010) Global forest resources assessment 2010: main report. Food and Agriculture Organization of the United Nations, Rome

  21. FAO and CIFOR (2005) Forests and floods. Food and Agriculture Organization of the United Nations, Rome and Center for International Forestry Research Bogor, Indonesia

  22. Ferraro PJ et al (2012) Forest figures: a review of ecosystem services valuation and policies in developing countries. Rev Environ Econ Policy 6:20–44

    Article  Google Scholar 

  23. Forster DL et al (1987) Soil erosion and water treatment costs. J Soil Water Conserv 42:349–352

    Google Scholar 

  24. Freeman AM III et al (2014) The measurement of environmental and resource values: theory and methods, 3rd edn. RFF Press, Taylor & Francis, Oxford

  25. Freeman J et al (2008) Statistical analysis of drinking water treatment plant costs, source water quality, and land cover characteristics. Trust for Public Land, San Francisco

    Google Scholar 

  26. Gartner T et al (eds) (2013) Natural infrastructure: investing in forested landscapes for source water protection in the United States. World Resources Institute, Washington

  27. Hewlett JD (1982) Principles of forest hydrology. University of Georgia Press, Athens

    Google Scholar 

  28. Holmes TP (1988) The offsite impact of soil erosion on the water treatment industry. Land Econ 64:356–366

    Article  Google Scholar 

  29. Huber PJ (1967) The behavior of maximum likelihood estimates under nonstandard conditions. In: Neyman LM (ed) Proceedings of the fifth Berkeley symposium on mathematical statistics and probability, vol 1. University of California Press, Berkeley, pp 221–233

  30. Hurley T, Mazumder A (2013) Spatial scale of land-use impacts on riverine drinking source water quality. Water Resour Res 49:1591–1601

    Article  Google Scholar 

  31. Kennedy P (2008) A guide to econometrics, 3rd edn. Blackwell, Malden

    Google Scholar 

  32. Kenny A (2006) Ecosystem services in the New York City watershed.

  33. Krishnaswamy J et al (2001) Spatial patterns of suspended sediment yields in a humid tropical watershed in Costa Rica. Hydrol Process 15:2237–2257

    Article  Google Scholar 

  34. Lembaga Air Perak (n.d.) Laporan tahunan 2010. Ipoh, Perak, Malaysia

  35. McConnell KE, Bockstael NE (2005) Valuing the environment as a factor of production. In: Mäler KG, Vincent JR (eds) Handbook of environmental economics, vol 2. North-Holland, Amsterdam, pp 621–669

    Google Scholar 

  36. McDonald R, Shemie D (2014) Urban water blueprint: mapping conservation solutions to the global water challenge. The Nature Conservancy, Washington

    Google Scholar 

  37. Mohd Akbar HJ, Rusnah R (2004) Water resources management for domestic and industrial needs. In: Abdul Rahim N (ed) Water: forestry and land use perspectives, Technical Documents in Hydrology No. 70, International Hydrological Programme. UNESCO, Paris, pp 1–15

  38. Mohd Nordin A et al (2000) The effects of raw water turbidity in water treatment process at Sg. Kampar treatment plant. In: Proceedings of environmental seminar, School of Chemical Engineering. Universiti Sains Malaysia, Penang, pp 46–51

  39. Moore WB, McCarl BA (1987) Off-site costs of soil erosion: a case study in the Willamette Valley. West J Agric Econ 12:42–49

    Google Scholar 

  40. Moulton BR (1986) Random group effects and the precision of regression estimates. J Econom 32:385–397

    Article  Google Scholar 

  41. Murray C, Forster L (2001) A study of pesticide use, farming practices, and community drinking water treatment costs in the Maumee and Great Lakes Basins. Rep. Series AEDE-RP-0013-01, Agricultural, Environmental, and Development Economics, Ohio State University, Columbia, Ohio

  42. Ogden FL et al (2013) Effect of land cover and use on dry season river runoff runoff efficiency and peak storm runoff in the seasonal tropics of Central Panama. Water Resour Res 49: doi:10.1002/2013WR013956

  43. Pattanayak SK (2004) Valuing watershed services: concepts and empirics from southeast Asia. Agric Ecosyst Environ 104:171–184

    Article  Google Scholar 

  44. Pattanayak SK, Wendland KJ (2007) Nature’s care: diarrhea, watershed protection and biodiversity conservation in Flores, Indonesia. Biodivers Conserv 16:2801–2819

    Article  Google Scholar 

  45. Pattanayak SK, Kramer RA (2001) Worth of watersheds. Environ Dev Econ 6:123–146

    Article  Google Scholar 

  46. Piper S (2003) Impact of water quality on municipal water price and residential water demand and implications for water supply benefits. Water Resour Res 39: doi:10.1029/2002WR001592

  47. Postel SL, Thompson BH Jr (2005) Watershed protection: capturing the benefits of nature’s water supply services. Nat Resour Forum 29:98–108

    Article  Google Scholar 

  48. Sagoff M (2002) On the value of natural ecosystems. Polit Life Sci 21:19–25

    Google Scholar 

  49. Schwabe KA et al (2014) Creation of Malaysia’s Royal Belum State Park: a case study of conservation in a developing country. J Environ Dev. doi:10.1177/1070496514551173

  50. Sklenar K et al (2012) Source water protection vision and roadmap. PDF Report #4176a, Water Research Foundation, Denver, Colorado

  51. Spiller M et al (2013) Pollution source control by water utilities: characterisation and implications for water management. Water Environ J 27:177–186

    Article  Google Scholar 

  52. Sthiannopkao S et al (2007) Soil erosion and its impacts on water treatment in the northeastern provinces of Thailand. Environ Int 33:706–711

    Article  Google Scholar 

  53. Tan-Soo JS et al (2014) Econometric evidence on forest ecosystem services: deforestation and flooding in Malaysia. Environ Resour Econ 59. doi:10.1007/s10640-014-9834-4

  54. Vincent JR (2011) Valuing the environment as a production input. In: Haque AKE et al (eds) Environmental valuation in South Asia. Cambridge University Press, New Delhi, pp 36–77

    Google Scholar 

  55. Vincent JR, Hadi Y (1993) Malaysia. In: Council National Research (ed) Sustainable agriculture and the environment in the humid tropics. National Academy Press, Washington

  56. Vincent JR, Mohamed Ali R (2005) Managing natural wealth: environment and development in Malaysia. Resources for the Future, Washington

    Google Scholar 

  57. Vincent JR et al (2014) Tropical countries may be willing to pay more to protect their forests. Proc Natl Acad Sci USA 111:10113–10118

    Article  Google Scholar 

  58. White HL (1980) A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48:817–838

    Article  Google Scholar 

  59. Whittington D (2010) What have we learned from 20 years of stated preference research in less-developed countries? Annu Rev Resour Econ 2:209–236

    Article  Google Scholar 

  60. Wooldridge JM (1995) Score diagnostics for linear models estimated by two stage least squares. In: Maddala GS et al (eds) Advances in econometrics and quantitative economics: essays in honor of Professor CR, Rao edn. Blackwell, Oxford, pp 66–87

    Google Scholar 

  61. Wooldridge JM (2010) Econometric analysis of cross section and panel data, 2nd edn. MIT Press, Cambridge

  62. Zeger SL, Liang KY (1986) Longitudinal data analysis for discrete and continuous outcomes. Biometrics 42:121–130

    Article  Google Scholar 

Download references


This study was funded by the Global Environment Facility through the United Nations Development Programme (MAL/04/G31), with additional support from the Government of Malaysia (through the Ministry of Natural Resources and Environment and the Forest Research Institute Malaysia) and the Center for International Forestry Research (CIFOR). The cooperation of numerous Malaysian government agencies is gratefully acknowledged: the Forestry Department Peninsular Malaysia; the Departments of Agriculture, Drainage and Irrigation, and Statistics; the National Water Services Commission; and the Perak Water Board. We also thank David James and two anonymous reviewers for helpful comments.

Author information



Corresponding author

Correspondence to Jeffrey R. Vincent.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Vincent, J.R., Ahmad, I., Adnan, N. et al. Valuing Water Purification by Forests: An Analysis of Malaysian Panel Data. Environ Resource Econ 64, 59–80 (2016).

Download citation


  • Ecosystem service
  • Water purification
  • Forest
  • Malaysia
  • Valuation