1TECHNICAL MEMORANDUM Incorporating Market Volatility Into Financial Assurance Trust Fund Initial Balance Estimates to Adequately Protect Taxpayers From the Long-Term Operations & Maintenance Costs of PolyMet's NorthMet Project SDEIS Wastewater Treatment Systems Prepared By: Matthew Tyler Professional Consulting Forester PO BOX 511 Finland, MN, 55603. March 12, 2014 1.0 Introduction My name is Matthew Tyler. I am a professional Consulting Forester and a resident of Finland, Minnesota in Lake County. I am also homeowner and taxpayer in the State of Minnesota. Permitting decisions made in the State of Minnesota will impact the State's future financial liabilities and the health of the state budget, and consequently effect tax rates and the quantity and quality of services available to myself and the rest of the public. 1.1 Qualifications I have significant training and professional experience in mathematics, statistical modeling, and natural resources economics. I earned my bachelor's degree in Ecological Forestry with a minor in Conservation Biology from Prescott College in 2010. My undergraduate work included advanced courses in statistics, forest and population modeling, and natural resource economics, and as part of my studies I presented a poster at the Ecological Society of America scientific meeting in 2008. I have 8 years of field forestry, consulting forestry, and forest management experience. My duties as a consultant to private landowners include appraising timber values and conducting financial analysis of timber management decisions. Additionally, I have been a contract Firewise Coordinator for Lake County since June of 2011. In that capacity, I have assisted in estimating, budgeting, and managing more than $600,000 worth of hazardous fuel reduction project grants. I was also the primary author of a major revision to Lake County's Community Wildfire Protection Plan (CWPP). The CWPP revision included statistical and spatial analysis of historical wildfire events and fire weather conditions in the county, as well as creating a county-wide map of hazardous fuel conditions using the US Forest Service's Wild Fire Decision Support System (WFDSS) and associated software tools. Most recently, I conducted a detailed economic cost-benefit analysis on behalf of Lake County in support of a $3,000,000 metal-roof retrofit program grant proposal to the Federal Emergency Management Agency (FEMA). 2 2.0 Need for a Long-Term Treatment Fund In the NorthMet Project SDEIS, PolyMet Mining Inc. proposes to meet water quality standards post-closure by actively capturing and treating groundwater seepage from waste rock piles and tailings basins. Water quality modeling in the SDEIS suggests that active, mechanical collection and treatment systems are likely be needed for at least 500 years at the plant site and 200 years at mine site, although the SDEIS claims that it is uncertain how long mechanical treatment will be required. In any rate, "it is expected to be long term." (SDEIS December 2013, p. ES-11, ES-24, 5-7). This raises the question of how to pay for long-term active, mechanical treatment. Commodity metal prices are volatile, and mining companies often go bankrupt. To protect taxpayers from bearing the cost of long term treatment, a portion of the financial assurance required by the State of Minnesota should be set aside to create a non-refundable trust fund dedicated solely to paying for the operations and maintenance of long term treatment. Such a trust fund is referred to in the remainder of this document as the Long-Term Treatment Fund (LTTF), or simply the Fund, to differentiate it from other aspects of financial assurance used to fund mine site and plant closure or contingencies for catastrophic natural disasters or engineering failures, or other accidents. Although we do not examine these other aspect of financial assurance here, it is our belief that those aspects are also vitally important to properly protecting the taxpayer, and must also be examined in depth. The value of trust in any given year are subject to considerable variation in rates of return due to market volatility. Similarly, the annual costs of operations and maintenance in any given year are subject to considerable variation in annual cost inflation. In some industries, annual inflation is considerably more volatile than the consumer price index. While many studies have shown that investment funds can be modeled by their average rate of return over the long-term, this modeling assumption only holds when the portfolio is subject to a buy and hold strategy. When the portfolio must make regular pay-outs, investment strategy usually becomes more conservative, because the fund is much more sensitive to short-term market fluctuations. It is the purpose of this document to calculate the initial size of the Long-Term Treatment Fund that will be large enough to withstand market and cost variation over 500 of years. However, we also acknowledge that this approach is based on underlying assumptions that may not be tenable: "Similarly, the assumption that financial assurance instruments can be developed to ensure that funds will be available centuries from now is not logical. The State of Minnesota has existed for 155 years. The United States of America has existed for 237 years. The notion that a mining company and financial assurance instruments will be available to work on a mine site 500 years from now is not believable." (Tribal Cooperating Agencies Cumulative Effects Analysis, PolyMet SDEIS Appendix C, pg. 152) 3 3.0 Basic Fund Model The value of the Long-Term Treatment Fund at the end of any given year can be described as the value of the fund from the year before plus investment returns during the year, minus operations and maintenance (O&M) costs during the year. Furthermore, operations and maintenance costs are equal to the O&M costs from the prior year plus the O&M costs from the prior times the rate of cost inflation over the year. Mathematically, this can be represented as: F (t) = F (t-l) + [F (t-l) * R (t) ] - [C (t-l) + I (t) *C (t-l) ] = F (t-l) * (1 + R (t) ) - C (t-l) * (1+ I (t) ) (EQ 1) Where F (t) is the value of the fund at the end of year t F (t-l) is the value of the fund at the end of year t - 1 C (t) is the annual cost of PolyMet's post-closure O&M at the end of year t C (t-1) is the annual cost of PolyMet's post-closure O&M at the end of year t-1 R (t) is the rate of return of the fund during year t expressed as a percentage (e.g. 7%) I (t) is the rate of cost inflation during year t expressed as a percentage (e.g. 2%) For example, if the initial balance of the fund (F 0 ) is $100 million, the initial O&M cost (C 0 ) is $5 million, cost inflation (I 1 ) over the year is 2%, and fund return (R 1 ) over the year is 7%, then the value of the Fund (in millions) at the end of the first year is: F 1 = F (0) * (1 + R (1) ) - C (0) * (1+ I (1) ) = $100 * (1 + 0.07) - $5 * (1 + 0.02) = $101.9 And the balance of the fund at the end of year 2, assuming stable returns and inflation rates would be: F 2 = F (l) * (1 + R (2) ) - C (1) * (1+ I (2) ) = $101.9 * (1 + 0.07) - $5.1 * (1 + 0.02) = $103.83 Note that this is a simulation representation of the Fund balance at year t. For constant inflation and return rates, a closed-form equation for the fund balance at year t is available by solving a corresponding differential equation. However, for real world rates of inflation and investment return that vary randomly over time, a closed form equation is not available using conventional calculus, but instead requires solving a difficult stochastic differential equation. Due to this difficulty, I opted to use a simulation approach. 3.1 Fund Simulation with Variable Rates of Inflation and Return The Variable Inflation and Return Rate Fund model was designed to simulate 500 years of waste water capture and treatment, as that is the maximum length of time simulated in the SDEIS. The objective of the model was to estimate the probability of the fund being solvent after 500 years for varying combinations of initial fund balances and annual O&M costs. Solvency was defined as the fund balance being greater than that year's annual O&M cost. 4 For a given combination of initial fund balance and annual O&M cost, the fund solvency probability was calculated by running the model 50,000 times. Each model run consisted of starting with the initial fund balance and annual O&M cost, and then iterating Equation 1 500 times, each time with a different randomly chosen rate of return and inflation rate. The rates were randomly chosen using normal probability distributions derived from the empirical data in sections 4 and 5. fund balance was recorded at the end of that run, and then the simulation was run again. The solvency probability was then calculated by dividing the number of solvent runs by 100,000 - the total number of runs. The simulation was conducted using the R project for statistical computing (R Core Team 2013) using the source code in Appendix C. 4.0 Empirical Estimates of Trust Fund Rates of Return It was assumed that the Long-Term Treatment Fund would be managed by the Minnesota State Board of Investment (MSBI). The MSBI handles many different pension and trust funds for the State of Minnesota. Of these funds, the two most similar in purpose to the Long-Term Treatment Fund are the Permanent School Fund and the Environmental Trust Fund. Both of these funds are managed with the intent of making yearly payments (MSBI 2013). However, of the two, the Permanent School Trust Fund is more conservative and aimed at long-term growth. The law requires that the Permanent School Trust Fund’s principal remain inviolate, and the fund is managed somewhat more conservatively, with about 50% of the assets allocated to bonds. (MSBI 2013). In this regard, the Permanent School Fund is probably the best analogue for the Long- Term Treatment Fund, which needs to remain solvent for a very long time while making yearly payments. Annual rates of return for the Permanent School Fund between 1987 and 2013 were compiled using public reports by the State of Minnesota that were available online. The MSBI has published annual reports online on the performance of all its funds between 1998-2013 (MSBI 2013). Permanent School Fund returns between 1987-1997 were estimated from a graph in Nobles and Brooks (1998). Table 1: Minnesota Permanent School Fund Returns 1987-2013 Year Return Year Return 1987 6.9% 2001 -2.6% 1988 7.2% 2002 -6.2% 1989 16.1% 2003 6.3% 1990 6.3% 2004 10.2% 1991 10.3% 2005 6.5% 1992 15.1% 2006 4.8% 1993 6.1% 2007 13.4% 1994 -3.1% 2008 -3.6% 1995 14% 2009 -9.3% 1996 5.5% 2010 12.3% 1997 8.2% 2011 17% 1998 17.8% 2012 6.4% 1999 14% 2013 10.8% 2000 6.1% Mean 7.278% Standard Deviation 7.136% 5 These rates of return can be seen in Table 1 and Figure 1, along with their mean and standard deviation. Although the long-term arithmetic annual rate of return is 7.3%, it is also clear that returns were quite volatile from year to year. 5.0 Empirical Estimates of Cost Inflation Rates The most common way to model inflation for financial analysis is derive annual inflation from the Consumer Price Index (CPI), published by the US Bureau of Labor Statistics. However, specific industries sometimes have cost inflation patterns that differ from the CPI. It is therefore more appropriate to model industry-specific inflation using an industry-specific Producer Price Index (PPI) series published by the US Bureau of Labor Statistics. The PPI "measures the average change over time in the selling prices received by domestic producers for their output" but also includes some services. (US BLS, 2014). A search of the available PPI series did not reveal any series specific to water treatment services. However, a PPI series for "Support activities for metal mining" (Series ID PCU213114213114) from 1985-2013 was found. Annual percent inflation was calculated by dividing the Annual index value by the preceding year's annual index value. The resulting inflation rates can be seen in Table 2 and Figure 2, along with their mean and standard deviation. The arithmetic average annual rate of mining service cost inflation was 1.75%. Interestingly, this sequence was more than twice as volatile relative to the mean than the investment returns in Table 1, with a standard deviation of 3.929. The sequence is characterized by periods of inflation near zero punctuated by periods of rapid inflation and deflation. 6.0 Fitting Probability Distributions to Return and Inflation Rates In financial modeling, the stock return distributions are commonly modeled using a normal (e.g. Gaussian or bell curve) distribution. This practice dates back to at least the early 1900s, but was widely adopted in financial analysis after the publication of seminal works by Black-Scholes (1973) and Merton (1973). The use of the normal distribution to represent financial returns in Table 2: Support Activities for Metal Mining Producer Price Index Inflation 1986-2013 Year Inflation Year Inflation 1986 0.10% 2000 -9.32% 1987 0.50% 2001 7.89% 1988 1.89% 2002 -1.45% 1989 4.59% 2003 0.17% 1990 1.31% 2004 0.78% 1991 0.09% 2005 11.03% 1992 0.09% 2006 4.31% 1993 2.21% 2007 2.14% 1994 0.09% 2008 6.86% 1995 0.09% 2009 1.89% 1996 0.99% 2010 0% 1997 3.82% 2011 4.38% 1998 4.62% 2012 6.48% 1999 -1.64% 2013 -4.84% Mean 1.753% Standard Deviation 3.929% 6 financial analysis has since been institutionalized in derivative pricing, risk management, and other financial fields. A kernel density plot (e.g. smoothed histogram) of the investment return data was graphed and then overlaid with a plot of the fitted normal distribution (Figure 3). A similar plot was made for mining service inflation rates (Figure 4). The fits are relatively good given the small sample size. However, it is worth noting that the tails (i.e. high and low extremes) of the normal distribution are considerably smaller than the observed data, except for the right tail in Figure 3. This means that an extremely low or high rate will occur less frequently in the simulation than in the real world. This is not unexpected. There is considerable literature critiquing the over-reliance of financial models on normal distributions dating back to the 1960s (Mandelbrot 1962, 1963, 1967). In the wake of the 2008 financial collapse, this literature has regained popularity, and alternate techniques have been rapidly evolving (e.g. Frain 2009, Haas & Pigorsch. 2009). In light of these developments, we attempted to fit the data to an alternate distribution (the alpha-stable distribution) recommended by Mandelbrot (1967) and Frain (2009). Unfortunately, the sample sizes of the return and inflation data were too small to obtain a satisfactory fit. The normal distribution was therefore used for modeling purposes in this document. Consequently, this model likely tended to under-estimate the effects of volatility on the solvency of the fund. Put another way, the real world financial markets are considerably more risky than a well behaved computer model. Therefore, the results from this model should be viewed as a bare minimum estimate of the required initial balance of the long-term treatment fund. Actual requirements will likely need to be higher. 7 7.0 Long-Term Treatment Annual Operations & Maintenance Costs Although the NorthMet SDEIS gives estimates of $3.5-$6 million annually for post-closure operations and maintenance (O&M) (SDEIS, pg. 3-138), the source of these estimates is not documented. The purpose of this section is to use NorthMet project specifications, independent cost estimates, and standard scaling methods to make an independent estimate of annual NorthMet post-closure O&M costs. 7.1 NorthMet SDEIS Waste Water Treatment Plant Capacities PolyMet proposes to meet water quality standards post-closure by treating collected waste water at two reverse osmosis plants: one at the plant site, and one at the mine site. Although the main SDEIS document does not include specifications for these plants, treatment capacities can be found in the Water Modeling Data Packages. Mine Site "The total capacity of the long-term closure WWTF is assumed to be 600 gpm. The actual flow sent to the WWTF is typically half of the capacity or 300 gpm." (PolyMet 2013i pg 168). Plant Site "The treatment capacity of the WWTP used in the model is 2000 gpm during the first eight years, and 3500 gpm after that." (PolyMet 2013j, pg. 112) While reverse osmosis is not planned to be installed at the mine site until year 40, we assume it has been installed for the purposes of calculating long-term treatment funding. Similarly, we assume that the full capacity of the plant site reverse osmosis plant has been installed. Although some have questioned the accuracy of the modeling used to size these plants, it is not the purpose of this document to raise those questions. Rather, the purpose here is to estimate annual operations and maintenance costs assuming the specifications given in the SDEIS. Nonetheless, the results of the financial model presented here will certainly be sensitive to the accurate sizing of the reverse osmosis plants. Future long-term treatment fund calculations should be based on more detailed reverse osmosis plant specifications and cost estimates. 7.2 Published Reverse Osmosis Reference Cost Estimates Two preliminary cost estimates for reverse osmosis facilities for other projects in the region were identified to use as reference estimates. Coincidentally, both are also on the former LTV Steel property. The first reference cost estimate is for a 4000 gpm plant to treat water pumped from the Area 1 Pit and discharged to Second Creek. It was developed by Barr Engineering on behalf of Mesabi Mining in pursuit of a NPDES permit for the Mesabi Nugget Phase II project (Barr 2011). This estimate features three fairly similar alternatives based on an ultra-filtration to reverse osmosis 8 to evaporator to crystallization system. Although the cost-estimate worksheets feature a more detailed breakdown, the worksheets have significant mathematical errors, are based on fairly general cost estimates, and are largely conceptual in nature. The second reference cost estimate is for a smaller 600 gpm plant to treat water in Cell 1E of the LTV Tailings Basin that is believed to travel through the basin and discharge into Second Creek (Barr 2013). This estimate was developed by Barr Engineering on behalf of Cliffs Erie and PolyMet pursuant to a 2010 consent decree between the MPCA and Cliffs Erie. Although on the same site, this estimate is not part of the NorthMet project. This estimate features greensand pretreatment and reverse osmosis with three post treatment alternatives: (1) evaporation and crystallization, (2) vibratory enhanced sheer processing (VSOP), or (3) intermediate concentrate chemical precipitation (ICCP) and secondary reverse osmosis. This estimate was based on extensive pilot and bench testing and detailed equipment costing from manufactures. As such, it is probably the more accurate of the two reference cost estimates. Table 3 summarizes the capitol and annual O&M costs of the three alternatives for both reference reverse osmosis systems. Corrections were made to the annual O&M costs from Barr (2011). In the original tables, O&M cost totals were added incorrectly, and somehow managed not to include the estimates for ultra-filtration and reverse osmosis. The original cost estimate tables from Barr (2011) and Barr (2013) are in Appendices A & B. Table 3 includes breakdowns of the reference cost estimates to facilitate cost scaling. Annual O&M costs were broken down into fixed and variable components (AWWA 2007). Fixed costs are those that generally vary little from month to month or with monthly treatment volume, such as membrane and filter changes, routine equipment maintenance, and staffing. Variable costs are those expected to vary predictably with treatment volume, such as treatment chemicals, sludge hauling and disposal, and electricity to power pumps. Table 3: Published Reference Cost Estimates for Two Reverse Osmosis Treatment Systems Each with Three Alternatives. Reference Treatment Size Total O & M Fixed O&M Variable O&M Estimate Option gpm Capital Cost Cost ($) % Cap Cost ($) % Cap Cost ($) Barr 2013 Evap + Crystallization 600 $19,729,392 $1,492,573 7.57% $1,435,636 7.28% $56,939 Barr 2013 VSEP + Crystallization 600 $23,969,992 $1,579,447 6.59% $1,509,698 6.30% $69,750 Barr 2013 ICCP + SRO +Crystallization 600 $14,817,750 $1,901,833 12.83% $1,346,820 9.09% $555,014 Mean $1,657,951 Barr 2011 Alternative #1 4000 $40,462,964 $5,383,000 13.30% $4,592,000 11.35% $791,000 Barr 2011 Alternative #2 4000 $39,802,164 $4,613,000 11.59% $3,920,000 9.85% $693,000 Barr 2011 Alternative #3 4000 $40,628,164 $5,215,000 12.84% $4,466,000 10.99% $749,000 Mean $5,070,333 9 No attempt was made to evaluate whether the reference reverse osmosis systems were adequate to remove the pollutant concentrations predicted in the NorthMet SDEIS. The inclusion of these reference systems in this document does not constitute an endorsement of their adequacy for the NorthMet project. Rather, the reference systems were included to establish third-party cost baselines for financial assurance modeling. 7.3 Adjusting Reference Cost Estimates to PolyMet SDEIS Plant Specifications The design capacity of the Barr(2011) reference estimate is just slightly larger than the design capacity of the NorthMet tailings basin (plant site) reverse osmosis facility, while the capacity of the Barr (2013) reference estimate is exactly the same as the design capacity of the NorthMet mine site reverse osmosis facility. Therefore, the simplest estimate of total annual NorthMet post-closure O&M treatment costs is the mean O&M cost from Barr (2011) plus the mean O&M cost from Barr (2013). Alternatively, the annual O&M cost of the NorthMet Plant Site (tailings basin) reverse osmosis facility can be estimated by adjusting the more current and accurate Barr (2013) reference costs using a scaling factor based on relative capacity. Given the capital cost and capacity of a known treatment plant, AWWA (2007) state that the capital cost of a new plant can be estimated by: C b = C a * (S b / S a ) 0.80 (EQ 2) Where: C b = capital cost of new size plant C a = capital cost of known size plant S b = capacity of new plant, in gpm S a = capacity of known plant in gpm The exponent reflects economies of scale where per unit costs decrease as plant capacity grows. Total O&M costs for the new plant can then be estimated using: T o = C b * R f + (S b / S a ) * C vo (EQ 3) Where T o = total O&M cost of new plant R f = ratio of fixed O&M costs to capital costs for known plant C vo = variable O&M costs of known plant. and C b , S b, , S a are as above This is formula assumes that fixed O&M costs increase at roughly the same rate as capital costs, which is a common practice for estimating building maintenance costs (i.e. keeping depreciation at bay). In contrast, the formula assumes that variable O&M costs increase at the same rate as plant capacity. In other words, there is little economy of scale effect for 10 consumables; a 1000 gpm plant should use about twice as much treatment chemical and electricity as a 500 gpm plant. Using these formulas, we developed alternative capacity adjusted cost estimates for the 3,500 gpm NorthMet Plant Site (tailings basin) reverse osmosis facility based on the Barr (2013) reference costs. The results can be seen in Table 4. Although the Table 4 estimates are considerably higher than the reference costs in Barr (2011), they likely reflect Barr's advancements in process research and cost escalation since 2011. Table 4: Capacity Adjusted Cost Estimates Based on Barr (2013) for Three Alternative Systems for the 3500 gpm NorthMet Tailings Basin (Plant Site) Reverse Osmosis Facility Alternative Capital Cost Fixed O&M Cost Variable O&M Cost Total O&M Cost Evap + Crystalization $80,881,116 $5,885,422.21 $332,142 $6,217,565 VSEP + Crystallization $98,265,558 $6,189,042.39 $406,876 $6,595,919 ICCP+SRO+Crystalization $60,745,722 $5,521,318.89 $3,237,583 $8,758,902 7.4 Lowest, Highest, and Mean Estimates for Post-Closure Total Annual O&M Costs Combining all available estimates from Tables 3 and 4, the lowest, highest, and mean post- closure annual O&M costs were calculated and are shown in Table 5. These estimates are considerably higher than those found in the PolyMet SDEIS. Note that the lowest total annual O&M cost in table 5 is higher than the highest cost estimate in the SDEIS. Table 5: High, Mean, and Low Estimates for NorthMet Post-Closure Total Annual O&M Costs Mine Site RO Facility Plant Site RO Facility Reference Cost Alternative Cost Reference Cost Alternative Cost Total O&M Cost Low Barr 2013 Evap+Crystal $1,492,573 Barr2011 Altern. #2 $4,613,000 $6,105,573 Mean -- -- 1 $1,657,951 -- -- 2 $6,130,564 $7,788,515 High Barr 2013 ICCP+SRO $1,901,833 Table 4 ICCP+SRO $8,758,902 $10,660,735 1 Mean of three cost estimates from Barr (2013). 2 Mean of three cost estimates from Barr (2011) and three cost estimates in Table 4. 7.5 Third Party Adjustment of O&M Cost Estimates According to Tribal Natural Resources Agencies and the EPA, financial assurances are typically adjusted to reflect the cost of the government hiring a third party contractor should the mining company go bankrupt. Generally, it costs about twice as much for the government to hire a contractor for cleanup than it does for the mining company to conduct cleanup. (Margaret Watkins, Grand Portage Natural Resources Dept., pers. Communication). However, these cost estimates already contain considerable contingencies (30-40%) for cost overruns. We therefore applied a more conservative third-party multiplier of 1.75 to the total O&M cost estimates. The third-party adjusted values can be seen in Table 6. 11 8.0 Long Term Treatment Fund Simulation Results Having obtained reasonable high and low estimates for post-closure NorthMet O&M annual costs and statistical distributions for trust fund returns and mining service inflation, we conducted the simulations outlined in Section 3. The input parameters and simulation results are displayed in Table 6. Figures 5 displays the probability of fund solvency after 500 years as a function of initial fund balance for each estimate of annual O&M costs in Table 6. Table 6: Evaluation of Minimum Initial Balance Requirements for Solvency of NorthMet Long Term Treatment Fund After 500 Years Using Stochastic Investment Returns & Inflation Rates for Various Annual Costs O&M Cost Third-Party Simulation Minimum LTTF Set Aside Type Estimate O&M Length p > 0.95* p > 0.99* Low $6,105,573 $10,684,753 500 $333,000,000 $404,000,000 Mean $7,788,515 $13,629,901 500 $423,000,000 $513,000,000 High $10,660,735 $18,656,286 500 $579,000,000 $694,000,000 * Probability of fund solvency is greater than p-value, (e.g. 95% or 99%) 12 9.0 Conclusions The annual long-term operating and maintenance post-closure cost estimates in the PolyMet SDEIS are significantly lower than the lowest estimates calculated from published estimates for similar reverse osmosis plants on the LTV property. This is a significant problem. Independent estimates of long-term annual O&M post-closure costs range from $6.1 to $10.6 million. Lacking specific, detailed treatment water treatment plant plans specific to the NorthMet Project, there is still considerable uncertainty about annual long-term treatment costs. Stochastic simulation of volatile investment returns and inflation rates show that the minimum beginning balance for the Long Term Treatment Fund required to ensure water treatment for 500 years after closure is $333 million, and quite possibly $500 million or more. The Long Term Treatment Fund set aside is above and beyond the estimated $50-$200 million mine closure and reclamation cost estimates in the SDEIS (SDEIS, pg. 3-138). The SDEIS also fails to include an estimate of a financial assurance funds for accidents, natural disasters, and other unforeseen events. Not including additional contingency funds, the total financial assurance package should be at least $383-$700 million. 10.0 Recommendations • The annual post-closure operations and maintenance costs presented in the SDEIS should be rejected, as they are inconsistent with independent estimates of reverse osmosis operating and maintenance costs. • Because the accuracy of long-term treatment trust fund requirements depends considerably on the accuracy of annual post-closure costs, the NorthMet FEIS must include detailed specifications and cost estimates for the post-closure treatment operations and maintenance. • The NorthMet FEIS and permit applications should include financial assurance trust funds requirements for long-term treatment based on probabilistic analysis of investment return and inflation volatility similar to that presented here. • Deterministic estimates of long-term treatment trust fund requirements are typically 40-60% lower than those presented here, and should not be used in the FEIS or permits because the fund is much more likely to go bankrupt over time. • Future probabilistic financial analyses should ideally use alpha-stable or other non- normal distributions. • Future probabilistic financial analyses should use longer investment return and inflation rate time series to improve the fit of statistical distributions. 13 References American Water Works Association (AWWA), 2007. Reverse Osmosis and Nanofiltration, 2nd ed., Manual of Water Supply Practices M46. Barr Engineering Co. 2011. Area 1 Pit Water Treatment Evaluation in Support of the Nondegradation Analysis. Mesabi Nugget Phase II Project. Prepared for Steel Dynamics, Inc., Mesabi Mining, LLC. June 2011. Barr Engineering Co. 2013. Reverse Osmosis Pilot Test Report. SD026 Active Treatment Evaluation. Prepared for Cliffs Erie LLC, and PolyMet Mining Inc. June 2013. Black, Fischer and Myron Scholes. 1973. The Pricing of Options and Corporate Liabilities. The Journal of Political Economy. 81(3):637-654. Frain, John C. 2009. Studies on the Application of the a-stable Distribution in Economics. Ph.D. Thesis. University of Dublin: Ireland. Haas, Markus, and Christian Pigorsch. 2009. Financial Economics, Fat-Tailed Distributions. Encyclopedia of Complexity and Systems Science 2009, pp 3404-3435 Mandelbrot, B. B. 1962. The variation of certain speculative prices. Technical Report NC-87, IBM Research Report. Mandelbrot, B. B. 1963. The Variation of Certain Speculative Prices, Chapter 15, Cootner (1964b), pp. 307–332. M.I.T. Press. Mandelbrot, B. B. 1967. The variation of the prices of cotton, wheat, and railroad stocks, and of some financial rates. The Journal of Business 40, 393–413. Merton, Robert C. 1973. Theory of rational option pricing. The Bell Journal of Economics and Management Science. 4(1):141-183. Minnesota State Board of Investment (MSBI). 2013. Annual Report 2013. Minnesota State Board of Investment. St. Paul, MN. http://www.sbi.state.mn.us Nobles, James and Roger Brooks. 1998. School Trust Land: A Program Evaluation Report. Minnesota Office of the Legislative Auditor. Report #98-05. St. Paul, MN. PolyMet Mining (PolyMet). 2013i. NorthMet Project Water Modeling Data Package, Volume 1 - Mine Site, Version 12. March 8, 2013. PolyMet Mining (PolyMet). 2013j. NorthMet Project Water Modeling Data Package, Volume 2 - Plant Site, Version 9. March 1, 2013. R Core Team (2013). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. http://www.R-project.org/. 14 Appendix A: Reprints of Tables 23-25 from Barr (2013) 15 16 17 Appendix B: Reprints of Tables 3.1-3.3 from Barr (2011) 18 19 20 Appendix C: R Source Code for Probabilistic Long-Term Treatment Trust Fund Model library(MASS) # Initiate Variables reps <- 50000 # Number of Model Runs lowfa <- 100 # Low Bound of Initial Trust Fund Balance highfa <- 800 # High Bound of Initial Trust Fund Balance startannualcost <- 18.656286 # Annual O&M Cost in 2013 dollars # Initiate normal distribution parameters # Inflation infl.mu <- 1.753 infl.sd <- 3.929 # Returns return.mu <- 7.278 return.sd <- 7.136 # Initiate model parameters AssureVals <- seq(from=lowfa, to=highfa, by=1) pvals <- rep(0,length(AssureVals)) # Main loop for(fa in AssureVals ) { AnnualCost <- rep(startannualcost,reps) Fund <- rep(fa,reps) for(yr in c(1:500)) { # Begin loop code Fund <- Fund - AnnualCost ann.inflat <- (1 + 0.01 * rnorm(reps, mean = infl.mu, sd = infl.sd)) # Fail safe for out of bounds probability densities ann.inflat[ ann.inflat <=0] <- 1 AnnualCost <- AnnualCost * ann.inflat ann.return <- (1 + 0.01 * rnorm(reps, mean = return.mu, sd = return.sd)) ann.return [ ann.return <= 0] <- 1 Fund <- Fund * ann.return } pvals[ AssureVals == fa ] <- (sum(Fund > 0) / reps) } # Graph the Results dev.new(width=6.5, height=6.5) par(mgp=c(2, 0.75, 0)) plot(pvals ~ AssureVals, col="white", xlab= "Financial Assurance Value (in $ Millions)", ylab="Probability Fund > Cost at yr. 500" , ylim=c(0,1.1), main="Figure 6: Probability of Long Term Treatment Fund\nSolvency at Year 500", cex.main=1, cex.lab = 1, cex.axis = 1) abline(h=c(0.95), lwd = 2, col="gray") lines(pvals ~ AssureVals, lwd = 4, lty="dashed" ) title(sub = "Gray horizontal line is at 95% probability level", line = 3.5, cex.sub =0.75) # Find the minimum amount of money needed for a given probability of solvency min(AssureVals[pvals>0.95]) min(AssureVals[pvals>0.99])