Market Analysis, with Projected. Performance of a Slot-Machine. Facility at Plainridge Racecourse

Transcription

1 Market Analysis, with Projected Performance of a Slot-Machine Facility at Plainridge Racecourse May 28, Jason Street, Arlington, Massachusetts Telephone: Fax: cummingsw@aol.com

2 Contents Executive Summary i 1. Introduction 1 2. The Lay of the Land 3 3. Methodology 9 4. Assumptions Projections Conclusions 23 Exhibits 24 Appendix A: Details of the Gravity-Model Methodology ( includes Bibliography )

3 List of Exhibits Exhibit (following the Executive Summary) A Summary of Projections -- FY2013$ B Summary of Projections -- Then-Year $ (following the main body) 1 Gaming Facilities in New England -- Existing and Proposed 2 Detail for Southeastern Massachusetts 3 Recent Performance of Gaming Devices at Tracks and Casinos Nearby 4 Long-Term Trends at Nearest "Mega-Casinos" 5 Long-Term Trends at VLT Facilities in Upstate New York 6 Portion of Model Inputs 7 Gaming-Device "Power Ratings" 8 Projected Slot-Machine Gaming Revenues (series) 9 Total Consumer Spending on Slots & VLT, with Net Exports (series) 10 Projected Annual Slot (and VLT) Spending per Adult 11 Summary of Projections -- FY2013$ 12 Summary of Projections -- Then-Year $ 13 Cummings Projections Compared to Innovation Group's

4 Market Analysis, with Projected Performance of a Slot-Machine Facility at Plainridge Racecourse Executive Summary Background / Introduction Will Cummings, the author of this report, has been retained to conduct an analysis of the gaming market in New England and develop projections for the likely gaming revenues of a new slot-machine facility proposed for Plainridge Racecourse in the town of Plainville, Massachusetts. The Racecourse lies just off Interstate 495 not far from its intersection with I-95. This site enjoys excellent access from all parts of New England. Exhibits 1 and 2 in the main body of the report present maps of the area. I have developed projections for this casino based upon the performance of the existing gaming facilities in the Northeast, as well as those most comparable elsewhere around the country, by using a gravity-model methodology that is described in more detail in Appendix A. This methodology relates the numbers of people who live at various distances from each gaming facility to their patronage at each such facility based on the experience elsewhere, using distance (or more accurately, travel time) and size as its most salient variables. These types of models have been employed with much success in a multitude of other markets across North America. I have developed projections for Plainridge under a variety of assumptions regarding the location of the full-scale casinos which may be authorized in Massachusetts, including in particular a Peak Demand scenario in which (in its early year[s]) Plainridge faces no competition within the

5 ii State, and ultimately, a Full Competition scenario which assumes 3 full casinos at Suffolk Downs, Springfield (downtown), and Taunton. Assumptions My projections assume: o The Plainridge slot-machine facility will be comparable to those of existing race track casinos in the Northeast in terms of access, appearance, spaciousness and amenities. I have assumed that micro-access with respect to ingress and egress will be good. I have assumed the parking structure that is now almost complete, but no hotel on site. o I have examined multiple scenarios regarding competition: 1. A Peak-Demand scenario (which I would expect to extend for one to two years) in which Plainridge would be the only new gaming facility in Southern New England; and then 2. A range of scenarios in which one to three full casinos, on scales close to those of Foxwoods and Mohegan Sun, are put into operation at Suffolk Downs, Everett, Milford, Taunton, New Bedford, Palmer, and/or Springfield, Massachusetts. I have assumed 4,000 slots at each of these facilities, compared to 6,000± currently at each of the casinos in Connecticut, and 4,750 VLTs at Twin River in Rhode Island. 1 o The performance of the Plainridge facility and the underlying propensity to spend of the population surrounding it will be similar to those of existing facilities in comparable situations. I have specifically assumed average annual spending of $684 per distanceadjusted adult, which corresponds to a power rating of 95. o This assumption applies to a time of stabilized operations, which is typically one to three years down the road from the opening of a new gaming facility, and assumes industry-standard patterns of investment in bricks and mortar and in player rewards. Given that a substantial change in the competitive landscape can be foreseen, these patterns (and in fact the entire concept of stabilized operations ) may not be appropriate here. 1 With modern multi-denomination (and in many cases, multi-game) machines, fewer devices are needed to provide equivalent or better customer service, and most casinos are reducing their numbers of machines accordingly, including Foxwoods and Mohegan Sun, though not (yet) Twin River.

6 iii o The existing casinos of Connecticut continue to operate largely as they do today, while the VLT facilities in Rhode Island add table games as planned. o No other new gaming facilities are developed in Southern New England or nearby New York. Projected Performance Based on these assumptions, I took the detailed model described (in part) in the Appendix, calculated the numbers of distance-adjusted adults likely to patronize each facility, and applied the appropriate rates of spending for each. A summary of the most salient results is presented in Exhibit A. These are reported here as of stabilized operations and in terms of FY2013 dollars. Exhibit B translates these into then-year dollars for the first five years of operation assuming 2% annual escalation and a first-year startup factor of -15%. The projected numbers of visitors (in thousands) and slot-machine spending per visitor are also indicated on each of these exhibits. Over these five years I have assumed that (i) Plainridge is the only gaming facility in Massachusetts in 2015 and 2016; (ii) full casinos open in Springfield and at Suffolk Downs in 2017; and (iii) a casino opens in Taunton, as proposed by the Wampanoag Tribe, in The most representative projections for what I call the Peak Demand and Full Competition scenarios (2016 and 2018, respectively) are indicated by shading in light yellow. My analyses and projections are based upon the assumptions described herein. Some of these assumptions will inevitably not materialize, and unanticipated events and circumstances will occur. The actual results will therefore vary from my projections, and such variations may be material.

7 Exhibit A: Projected Total Gaming Revenues For a Slot-Machine Gaming Facility at Plainridge (FY2013$) Plainridge "Peak Demand" ("as if" in FY2013) Slot Win ($/day) Impact ($/vis) # Visitors (Cumulative) (000) (No New Gaming Elsewhere) $196.4 mn ($431) $84.5 2,323 (except tables at Twin River) Two Mass. Casinos open: Springfield and... Region A: Suffolk Downs (4,000 slots) $156.1 mn ($342) -21% $80.1 1,948 Everett (4,000 slots) $153.4 mn ($336) -22% $79.8 1,922 Milford (4,000 slots) $150.3 mn ($330) -23% $79.5 1,891 Region C: Taunton (4,000 slots) $151.1 mn ($331) -23% $79.6 1,899 New Bedford (4,000 slots) $167.4 mn ($367) -15% $81.3 2,058 Three Mass. Casinos open: "Light" competition (Suffolk, New Bedford & Sprgfield) $144.6 mn ($317) -26% $78.9 1,834 "Full" competition (Suffolk, Taunton & Springfield) $129.8 mn ($285) -34% $77.2 1,681 "Intense" competition (Milford, Taunton & Palmer) $115.5 mn ($253) -41% $75.7 1,526 Notes: All scenarios assume: 51% retention rate, therefore reasonable (though modest) CapEx and player rewards. Good micro-access. 24/7 operation. 1,250 machines, including electronic table games. Slot "Power Rating" (i.e., "everything else") = 95(-), similar to comparable facilities. All projections are as of "stabilized operations," and assume standard patterns of investment in facilities, marketing and player rewards. These may not be optimum here. Table games at Twin River and Newport. No gaming in NH.

8 Exhibit B: Projected Total Gaming Revenues For a Slot-Machine Gaming Facility at Plainridge (Then-Year $) Projections in FY2013$: Slot Win ($/day) ($/vis) # Visitors (000) (No New Gaming Elsewhere) $196.4 mn ($431) $84.5 2,323 (except tables at Twin River) Springfield + Suffolk Downs $156.1 mn ($342) $80.1 1,948 (assumed as of 1/1/2017) Springfield + Suffolk + Taunton $129.8 mn ($285) $77.2 1,681 (assumed as of 1/1/2018) Projections in Then-Year Dollars: (2.0%) Slot Win ($/day) ($/vis) # Visitors Year Base $ Escal'n Start-Up (000) 2015 $ $175.4 mn ($385) $85.4 2, $ $200.0 mn ($438) $89.4 2, $ $170.6 mn ($374) $87.6 1, $ $144.8 mn ($317) $86.1 1, $ $147.7 mn ($324) $87.8 1,681 Notes: All scenarios assume: 51% retention rate, therefore reasonable (though modest) CapEx and player rewards. Good micro-access. 24/7 operation. 1,250 machines, including electronic table games. Slot "Power Rating" (i.e., "everything else") = 95(-), similar to comparable facilities. All projections are as of "stabilized operations," and assume standard patterns of investment in facilities, marketing and player rewards. These may not be optimum here. Table games at Twin River and Newport. No gaming in NH.

9 Market Analysis, with Projected Performance of a Slot-Machine Facility at Plainridge Racecourse 1. Introduction The Commonwealth of Massachusetts recently authorized casino gaming in the form of (up to) three full-scale destination resort casinos (Category 1 licensees) and one smaller slot parlor (a Category 2 license, limited to 1,250 slot machines). Plainridge Racecourse, located in the Town of Plainville, Massachusetts, is in the process of applying for that Category 2 license. The Town has retained a team led by to assess the likely performance of the facility and its likely impacts on the Town. In this report, Will Cummings assesses the current state of the market for slot-machine gaming in New England and develops projections for likely gaming win and visitation to the proposed slot facility. In addition to their value as information in their own right, these projections also provide inputs to other members of the Cummings Team who assess, in separate reports, the likely social impacts of the new gaming facility, its effects on Town services and budget, and its economic impacts. Plainridge Racecourse is located at the intersection of US-1 and Interstate 495, not far from the intersection of the latter with Interstate 95. I-95 is the main artery between Boston and Providence, Rhode Island, and connects through Southern Connecticut to New York City. I-495 is the outer ring road around Greater Boston, and connects Cape Cod to the Massachusetts Turnpike, then circles around to the northern suburbs of Boston and connects with all the major routes into Southern New Hampshire, including I-95 again, which then proceeds to Maine. The site therefore enjoys excellent access from all parts of New England.

10 2 This report describes my analyses and conclusions. Section 2 presents background information regarding U.S. casinos in general, the competitive environment in New England in particular, the performance of slot machines and table games at the facilities that currently serve Massachusetts, New York and New England and those most comparable elsewhere, and recent trends in such gaming revenues. Section 3 describes my analyses of existing markets for gaming across the U.S., and explains my methodology for analyzing and projecting such revenues. (I discuss this methodology in greater detail in an Appendix.) Section 4 describes the key assumptions underlying my projections, which are then developed in Section 5. Section 6 presents a summary of my conclusions.

11 3 2. The Lay of the Land Casino gaming has proliferated across North America over the past twenty-four years. With the most recent addition of Massachusetts, eighteen states now authorize full-scale casino gambling on a commercial scale without limiting it to Indian tribes. In fifteen other states, full-scale casino operations are conducted only by Native Americans, but in at least nine of them, including Connecticut, California and Florida, these casinos are quite substantial. Seven of the 33 casino jurisdictions also authorize slot machines (or slot-like video lottery terminals, known as VLTs ) at their race tracks, 2 and two states have gaming devices at their race tracks only (i.e., without full casinos). A table which summarizes this information is presented on the following page. Whether at full-scale casinos, at race tracks, or at additional locations in eight other states, the public s appetite for gambling at slot machines is immense. They now constitute up to 90% of total revenues at most diversified casinos. As described in Appendix A, my analyses indicate that the average adult who lives with convenient access to a standard facility with slot machines spends roughly $720 per year on them (approximately 1.25% of personal income). A map which depicts the existing casinos in and near New England is presented in Exhibit 1. In this and the following maps, the black squares indicate existing casinos and the open squares indicate the prospective facilities that are assumed or considered as competition in this report. 2 That is, slots only, but not full casinos. In some of the other states that authorize casinos, full race track gaming facilities feature prominently in the mix (and in some cases were the original venues for gaming of any kind in the state). Examples include Delaware, Iowa, Pennsylvania, and West Virginia, which typically started with slots at tracks and later added table games and/or additional, non-track casino locations. In Delaware, however, gaming facilities are still limited to race tracks by statute.

12 4 State-Regulated Indian Casinos Slots at Tracks Only Casinos * Only * Colorado Arizona Arkansas 3 Delaware 1 California Rhode Island Illinois 4 Connecticut Indiana * Florida * Iowa Idaho Kansas Minnesota Louisiana * 4 Montana 4 Maine New York * Massachusetts New Mexico * 4 Michigan North Dakota Mississippi Oklahoma * Missouri Oregon 4 Nevada 4 Washington New Jersey Wisconsin Ohio * Wyoming Pennsylvania 5 South Dakota 4 West Virginia 4 * Indicates states with slots at tracks as well as / distinct from full casinos. 3 4 Arkansas s machines are technically limited to games of skill, such as video poker and blackjack. Louisiana, Montana, Nevada, New Mexico, Oregon, South Dakota, West Virginia and Illinois also authorize slots or slot-like VLTs at widespread bars, taverns, and/or fraternal establishments. In New Mexico, these are not economically significant, but most of the other widespread states, they are. Illinois passed such legislation in 2011, and has just begun to distribute these devices. 5 Pennsylvania began with slots at tracks, but will ultimately have five substantial standalone (i.e., non-track) gaming facilities. Four have opened so far, at Pocono (Mount Airy), Bethlehem (Sands), Pittsburgh (The Rivers), and Philadelphia (SugarHouse). A smaller facility just opened in Valley Forge. Pennsylvania authorized table games in 2010, so these are all now full casinos.

13 5 Detail for the immediate area is presented in Exhibit 2. Again, existing casinos are depicted by black squares, and locations proposed for new ones are shown as open squares. Plainridge Racecourse, at Plainville, MA, enjoys excellent access from the east and west via Interstate 495, and from the north and south via nearby Interstate 95. Plainridge currently provides live harness racing on a summer schedule, and wagering on simulcast racing from other tracks (which provides the bulk of its revenues) year-round. The top portion of Exhibit 3 presents recent statistics for the performance of slot machines at the six existing gaming facilities in New England and the two closest VLT facilities in New York. Five of these facilities are of monstrous size, two are highly urban, and three are rather rural. Several of these facilities (among both the urban and the rural) are somewhat difficult to access, so performance across all these casinos is highly varied. The power rating statistic that I present in the final column measures the performance of each facility in terms of its success in attracting spending from the surrounding population based on a gravity-model analysis that I describe later in this report. In brief, it is based upon the revenues of each facility compared to its size and accessibility to the surrounding population versus its competitors. The bottom portion of Exhibit 3 presents statistics for several of the gaming facilities in Delaware, Pennsylvania, and West Virginia that I believe are qualitatively very similar to the project proposed for Plainville, though they vary somewhat in size and urban vs. rural character. I am not concerned by the recent declines in business at many of these facilities, as these declines have been due to the opening of new competition in Maryland and Ohio, affecting Delaware and Pennsylvania/West Virginia, respectively.

14 6 I would observe at this point that aside from the Indian casinos in Connecticut, all of the other gaming facilities in this exhibit labor under tax rates (and other financial burdens, such as purse allocations for horsemen) that are higher than those which will be levied at Plainridge. In Pennsylvania, Delaware, and West Virginia, those burdens are only slightly higher, but in Rhode Island, the effective tax rate is much higher. As described in this Appendix A, this has significant adverse impacts on casino performance. Results at Plainridge should therefore be more similar to the comparable facilities listed here than to its nearest competitors in Rhode Island. (The gravity-model analysis which I describe in the following sections takes not only capacity, access, and demographics, but also tax rates into account in developing my projections.) My perspective has to this point been rather static, and will be so again later: what is the performance now, or if at Plainridge, how it would have done over the twelve months just ended ( FY ). Given, however, the severity of the recession from which we shall hopefully continue to emerge, it is reasonable to ask whether recent results, in the Northeast or anywhere else, provide a reasonable basis from which to project the future. The recession severely battered casinos in Nevada, Atlantic City, Connecticut and the Chicago area (aggravated there by the introduction of a ban on smoking in 2008). Elsewhere, however, and particularly in the Northeast, most nondestination gaming facilities held up rather well. Recent statewide statistics for slot win which illustrate these trends include: 6 While the twelve months which ended March 31, 2013, are not technically FY2013 as most states define it (New York is an exception), I believe they are close enough to use that as my designation for this recent 12-month period.

15 7 Slot Gaming Revenues State FY2013 vs. FY Destination Markets (relatively remote from most customers): Connecticut % New Jersey % Locals Markets (relatively close to most customers): Rhode Island + 4.8% Upstate New York % Pennsylvania + 4.2% Iowa + 2.6% New Mexico + 2.1% (South) Florida tracks % 8 Customers clearly economized during the downturn by gambling closer to home rather than traveling longer distances to destination resorts. Locals-oriented regional casinos held up far better, and in many cases actually saw their revenues grow, through the course of the recession. Gross VLT revenues in New York State, for example, increased by 34% between FY2008 and FY2012 (excluding Resorts World at Aqueduct), and those in Rhode Island, despite serious financial troubles at Twin River, by 10%. I therefore believe that recent performance does indeed provide a 7 Comparisons are generally for my pseudo-fy2013 vs. actual FY2011, but NY compares actual fiscal years (ending March 31), and the NJ comparison is for calendar 2012 vs I have not included Maine because its second casino opened just last June and therefore skews its results. 8 Florida race track slot performance continues to benefit from a large reduction in the state tax rate enacted in I have excluded the casino at Miami Jai-Alai, which opened in mid-fy2012, from this calculation.

16 8 reasonable guide to that which we should expect in the near future barring yet another recession, of course. Longer-term trends are presented in the graphs of Exhibit 4 for the mega-casinos in Southern New England and the New York City area, and for a selection of the VLT facilities in Upstate New York in Exhibit 5. I have not prepared similar graphs for the casinos of New Jersey, Delaware, Pennsylvania, and West Virginia, because they are uninformatively erratic due to a continuing series of competing new casinos opening within each of those or neighboring jurisdictions. I believe, however, that these two exhibits demonstrate quite well my general thesis that local gaming facilities have held up well, while destination resorts have suffered Connecticut s in particular, due in no small part to the large locals establishments which now cramp their reach to the east and west.

17 9 3. Methodology In order to develop my projections for the new gaming facility at Plainridge, I first conducted detailed analyses of the performance of the existing casinos in the Northeast (and those most comparable in other states) in relation to the demographics of the market areas which surround them. Geography, by which I primarily mean the distribution of population, is the most important factor underlying the performance of gaming facilities, as it is indeed for the sales of most consumer goods and services. My methodology is described in detail in Appendix A. In brief, it is based on the number of adults residing at various distances from each gaming facility in an area, and the ratio of actual revenues obtained to such numbers of adults so distributed. I apply gravity models that incorporate data for various geographic subunits in each market such as its adult population, per capita income, urban/rural nature, and travel time to the nearest casino(s) and/or race-track gamingdevice facilities (or relevant group(s) of such facilities). 9 From these parameters, I estimate the distance-adjusted adult population of each market. This figure is intended to represent the effective market population as if the entire population resided within ten minutes of a gamingdevice facility. 10 In order to do this on a detailed basis, I conducted this analysis by zip code in most of New England and the neighboring portions of New York and New Jersey. In order to analyze 9 These are called gravity models because in their simplest form, they are similar to Newton s Law of Gravitation: the attraction of each competing facility is inversely proportional to the square of its distance from the relevant population. Because economists named Reilly and Huff pioneered their application to retail sales, they are now called Reilly or, more commonly, Huff models. 10 And also had per capita income of $38,846 (the U.S. average in 2009) and was urban in nature, i.e., part of a defined metropolitan statistical area. These relationships are based upon statistical analysis of these models and of survey data from several large casino markets. The ten-minute criterion is no special figure; it is simply a benchmark intended to represent convenient access.

18 10 the performance of the most relevant existing casinos, the model also covers portions of several other states, extending as far as Virginia to the south and Ohio to the west. For the areas most remote from existing casinos, I used counties as the basic elements of analysis. 11 A portion of the detailed gravity model is illustrated in Exhibit 6 (note that this exhibit extends over two pages, and presents just a small portion of the model). In addition to the demographic data pertaining to each zip code, a second set of inputs describes its distance from each of many current or potential gaming facilities or groups of such facilities: Foxwoods and Mohegan Sun in Connecticut, Twin River (Lincoln) and Newport Grand in Rhode Island, Bangor and the new casino at Oxford, Maine, and then potential locations in Massachusetts. Other pages cover Atlantic City, the gaming facilities in New York, Pennsylvania, Maryland, West Virginia, and many other casinos on the fringes of this region. The model takes the travel time from each geographic market segment to each of the competing gaming-device facilities, identifies the closest such facility, and based on the distance/travel time, estimates a distance-adjusted adult population for each market segment. Again, these figures represent the number of adults that would generate the estimated level of spending if they all lived within ten minutes of the facility. For the State as a whole (and for all the adjoining markets), these figures are lower than the actual adult population, because most people live more than ten minutes from such a facility. The model then distributes the distance-adjusted adult populations of each market segment across all the competing facilities, depending upon travel time and attractiveness. 11 Because I have used similar models to develop projections elsewhere, I have actually analyzed all of Pennsylvania, Delaware, and Maryland at the zip-code level, as well as most of the Midwest.

19 11 I have used such models for analysis and projections extensively in the Northeast, Midwest, West and many other markets across the country in a similar fashion. The results of these are summarized in Exhibit 7. This exhibit presents an index that I call a power rating for each of the facilities (or groups thereof) in most of the major markets of the U.S. 12 These power ratings represent annual spending on slot machines (and/or VLTs) per distance-adjusted adult compared to a representative Midwest Standard figure of $720 (the middle yellow bar in this exhibit). Annual spending averaging $792 per (distance-adjusted) adult would translate into a power rating of 110 (the upper yellow bar); $648 translates into a power rating of 90 (the lower yellow bar). (Additional detail is provided in Appendix A.) Exhibit 7 presents the broad range of markets in three groups: Northeastern U.S. and Florida in the first column, medium to large markets elsewhere in the second column (largely in the West and Midwest), and very rural markets in the third column. Rural facilities often do very 12 It may be helpful to consider these power ratings as a kind of extension of the fair share concept that is often used to compare different gaming facilities. If, for example, all the slot machines in a given market average $200 in win per day as a group, a facility at which they win $240 per day is said to do 120% of its fair share. One that wins $180 per machine per day is said to do just 90% of its fair share. The concepts behind my power ratings are similar, but include analyses of the surrounding demographics. If there are many people and few machines in an area (Chicago, for example), each machine should win a lot each day. In rural Iowa, on the other hand, there are many more machines and many fewer people. A facility with the same power rating in Iowa will win much less per machine per day than in Chicago, but will do equally well in attracting spending from the (smaller) surrounding population. The major omissions from this chart involve California and Arizona. In these two states (as in many others), tribal gaming operations rarely release revenue statistics. In addition, I have not had the opportunity to analyze the markets of Louisiana and Mississippi in great detail.

20 12 well. 13 More competitive markets also generally attract higher rates of spending, but as they may be either urban, rural, North, South, East or West, I have not devoted a separate column to them. It is obvious that many of the gaming facilities in the Northeast (the first column) perform poorly by this measure. This is due in large part to the relatively high tax rates imposed on most slots-at-tracks facilities there and in Florida, too (at first), at the bottom of that column. (The bottom of the middle column, by contrast, generally comprises old-style riverboats that have not yet been replaced by more spacious and micro-accessible modern facilities.) Under high-tax conditions, casino operators can invest only very modest amounts in bricks and mortar and in player rewards, and these are increasingly critical to most gaming operations today especially in competitive markets. There is, in fact, a strong correlation between slot performance and tax rate, or more precisely its converse, the retention rate which casinos are allowed to keep. 14 I discuss this relationship in more detail in the Appendix, but would simply emphasize again that the relatively reasonable tax rate in Massachusetts provides a reward structure that will likely result in much better performance at its new facilities (both full casino and slot parlor ) than that which is demonstrated by those of Rhode Island. Based upon the experience elsewhere and the specific assumptions that I describe in Section 4, the model applies the elasticities of spending versus distance and income to the population of all 13 As described in Appendix A, there are at least three reasons why rural facilities appear to do so well: (i) easier to get around, (ii) less to do, and (iii) survival bias in small markets, sometimes only the best survive. 14 In addition to taxes on gross revenues, at race track facilities (and some others) the retention rate also reflects the subtraction of mandatory purse payments to horsemen, breeders' funds, and miscellaneous social mandates.

21 13 the relevant zip codes. It calculates the market shares of each casino or group of casinos, and also incorporates the potential impacts of capacity constraints. The model then calculates the impacts of all these factors on the appropriate rate(s) of spending per adult per year from each zip code or county, and allocates that spending among all the facilities in the region in particular, the proposed slot-machine facility in Plainville.

22 14 4. Assumptions I have assumed for my projections that: o The gaming facility at Plainridge will be comparable to those of existing race track casinos in the Northeast in terms of access, appearance, spaciousness and amenities. I have assumed that micro-access with respect to ingress and egress will be good. I have assumed the parking structure that is now almost complete, but no hotel on site. o I have examined multiple scenarios regarding competition: 1. In the expectation that Massachusetts s slot facility will be licensed, built and put into operation well before any of its full-scale casinos, a Peak-Demand scenario (which I would expect to extend for one to two years) in which Plainridge would be the only new gaming facility in Southern New England; and then 2. A range of scenarios in which one to three full casinos, on scales close to those of Foxwoods and Mohegan Sun, are put into operation at Suffolk Downs, Everett, Milford, Taunton, New Bedford, Palmer, and/or Springfield, Massachusetts. I have assumed 4,000 slots at each of these facilities, compared to 6,000± currently at each of the casinos in Connecticut, and 4,750 VLTs at Twin River in Rhode Island. o The performance of the Plainridge facility and the underlying propensity to spend of the population surrounding it will be similar to those of existing facilities in comparable situations. I have specifically assumed average annual spending of $684 per distanceadjusted adult, which corresponds to a power rating of 95. o This assumption applies to a time of stabilized operations, which is typically one to three years down the road from the opening of a new gaming facility, and assumes industry-standard patterns of investment in bricks and mortar and in player rewards. Given that a substantial change in the competitive landscape can be foreseen, these patterns (and in fact the entire concept of stabilized operations ) may not be appropriate here. o The existing casinos of Connecticut continue to operate largely as they do today, while the VLT facilities in Rhode Island add table games as planned. o No other new gaming facilities are developed in Southern New England or nearby New York.

23 15 In addition to spending by local residents, in some cases my models include contributions from (long-distance) drive-by traffic, hotel guests, and/or seasonal residents. I do not believe that these will be significant here. Plainridge does not plan a hotel, and the area has no extraordinary number of seasonal residents. The site is immediately adjacent to Interstate 495, but with no hotel, no great visibility from the highway, and no plans for a travel plaza (nor even good parking for truckers), I believe the project minimally suited to attract casual travelers. In addition, the vast majority of passers-by are residents of New England, and therefore captured in the gravity-model analysis. Most of the minority traveling long-distance rather than locally are likely headed to or from Cape Cod, a destination (or origin) that is for the most part just one to two hours away. I believe these conditions are not conducive to any significant drive-by contribution to the business and have therefore added none to my projections. My projections are thus based solely on my gravity model of the region. I took the detailed model illustrated (in part) in Exhibit 4, calculated the numbers of distance-adjusted adults likely to patronize the new facilities I have assumed under each scenario, and applied the appropriate average rates of spending to each. The results are described below.

24 16 5. Projections I have developed projections under a variety of scenarios regarding the locations of the other casinos in Massachusetts. I describe only a selection of these scenarios here: A. Peak Demand assumes that the full casinos in Massachusetts (Category 1 licensees) have not yet opened. This situation will likely last for just one or two years; I have assumed the latter. B. Because the situation in Southeastern Massachusetts ( Region C ) remains clouded by issues surrounding an earmark for the Wampanoag Tribe (which the Massachusetts Gaming Commission has now allowed to lapse, but remains disputed by the Tribe), a casino in that region will likely take (much?) longer to develop than those approved in the two other zones into which the Commonwealth has been divided. I have therefore examined intermediate scenarios in which there are just two competing full casinos in Massachusetts. This situation will likely last for at least one year, but may extend much longer. 15 C. Finally, I examine scenarios in which a casino does get built in Southeastern Massachusetts, one of which I characterize as a reasonably likely full competition scenario. Under all these scenarios, I base my analyses and projections on the performance of facilities elsewhere in FY2013, and therefore calculate them initially in terms of FY2013 dollars. I then extrapolate to future years assuming normal growth, due to rising local population, incomes, and 15 Even if courts do not interfere with the Gaming Commission s process to license a full commercial casino in Region C, the Wampanoags, if and when they do ultimately obtain Tribal lands in that area, will remain eligible under the Indian Gaming Regulatory Act to build a casino on such lands. This will surely complicate efforts to finance any other casino authorized in that Region under the Commonwealth s current statutory process.

25 17 inflation, at 2% per year. There is, however, typically an initial transient (ramping up) of 5 to 15 percent as new gaming facilities work out their kinks. This is usually most severe in highly competitive markets, and I anticipate that will be the case here. Plainridge will have to develop its players list and rewards programs, which are critical to the success of most casinos today, in the face of what will likely be strenuous efforts by its existing competitors to retain their current players A. Peak Demand This scenario assumes that none of the full casinos of Massachusetts has opened. My projections for this scenario are presented in Exhibit 8-1 please pardon all the blank space, but I often find it helpful to use a single template that accommodates multiple scenarios in order to make comparisons among them easier. When running solo (in Massachusetts), I expect the 1,250 slot machines at Plainridge to generate $196.4 million in annual slot win in terms of FY2013, for an average of $430 per slot machine per day. At the projected rate of utilization, the facility will be packed on Friday and Saturday evenings, and sufficiently crowded at many other times that large numbers of customers will not be able to walk right in and sit down at their favorite machine. As a result, as footnoted in the exhibit (and similarly in many of those that follow), the actual power rating that I expect the facility to demonstrate will be somewhat less than the 95 that I assume as a benchmark. With no competition from other casinos in Massachusetts, the demand for slot-machine gaming will far exceed the supply of 1,250 machines to which Plainridge is limited by Massachusetts law.

26 18 B. Two Category 1 Casinos With two competing full casinos, my projections indicate that total win (and visitation) at Plainridge will likely be reduced by 15% to 23%. One such scenario is illustrated in Exhibit 8-2. This assumes a 4,000-slot casino at Suffolk Downs and one (as proposed by MGM) in downtown Springfield (each would also have table games, a hotel, and a plethora of restaurants and other amenities. I therefore assume the slot power ratings for these types of casinos to be in the range of 100 to 103, with the higher ratings in more rural areas. As I described above regarding Plainridge, however, crowded conditions will reduce the effective power ratings of these competing facilities.) Under these conditions, I project total annual slot win at Plainridge to be $156.1 million, an average of $342 per unit per day. Under alternative configurations of two casinos, my projections differ slightly: Full Casinos At Springfield and Everett Springfield and Milford Springfield and Taunton Springfield and New Bedford Total Annual Win at Plainridge $153.4 million $150.3 million $151.1 million $167.4 million I have included Springfield (downtown) in all these scenarios because the gaming revenues at Plainridge are not very sensitive to the precise location of the full casino in Western Massachusetts; if it is at the Eastern States Exposition in West Springfield, Plainridge will do just slightly better, and if at Palmer, as proposed by Mohegan Sun, just slightly worse.

27 19 As indicated in the table above, with only two full casinos in Massachusetts my projections for Plainridge are in fact not very sensitive to their precise locations in Eastern Mass. This is because, at reasonable sizes, they will not be sufficient to soak up any truly disproportionate share of the available demand. It may seem surprising that competing casinos at Milford or Taunton have so little incremental impact on Plainridge, but that is because they would leave the core of the Boston area very poorly-served. Despite its lower power rating / attractiveness, Plainridge s competitive disadvantage against them would be eased by the fact that most Bostonians (and residents of the North Shore) would find Plainridge almost equally accessible. The much better access that casinos at Suffolk Downs or Everett would have to many more people almost offsets their greater distance from Plainridge as a competitive factor. C. Three Category 1 Casinos With three full casinos in Massachusetts, supply comes more into balance with demand and I project that slot revenues at Plainridge would take another step down. In this situation, however, the precise location of those competing casinos becomes more important: Full Casinos At Suffolk, New Bedford & Springfield Suffolk, Taunton & Springfield Milford, Taunton & Palmer Total Annual Win at Plainridge $144.6 million $129.8 million $115.5 million I present details for the second option in Exhibit 8-3 (Suffolk, Taunton and Springfield). As it lies in the middle of this range, I would consider it the most representative Full Competition scenario, and have used it as such for the five-year projections I present in Exhibit 12.

28 20 Impacts on Net Exports of Gaming Services One prominent motivation for the legislation authorizing casinos in Massachusetts was the fact that residents of the Commonwealth currently spend nearly a billion dollars each year at casinos in other states (primarily Connecticut). Casinos in Massachusetts will recapture much of this spending, and ideally reverse the flow to bring significant out-of-state gaming dollars into the State. Such net exports, arising directly from the spending of the residents of other states or by means of substituting for the gaming services that Massachusetts residents currently import, provide real value to the State s economy far beyond recirculation of consumer spending within Massachusetts itself. My calculations for such net exports provide input to Civic Economics projections for the likely overall economic impacts of the Plainridge project that are described in a separate report. The gravity model s projections for such net exports are summarized for a selection of scenarios in Exhibits 9-0 through 9-3B. I start with what I call the null case in Exhibit 9-0. This is based upon my analysis of the current day, but adds the impacts which I estimate the addition of table games to Twin River (and Newport Grand) in Rhode Island will have on Massachusetts residents spending at their facilities. (I am not concerned with the revenues of the table games themselves, but simply with the modest increase, on the order of 6%, that I believe they will induce in VLT play in Rhode Island.) I estimate that following that development, anticipated for July of this year, Massachusetts residents will be spending roughly $756 million on slot machines and VLTs in other states. (This is less than the billion-dollar figure cited above because that larger figure includes spending on table games as well as slot machines.) With no casinos of its own (yet), Massachusetts net exports of gaming services will be a negative $756 million, i.e., we will be importing such services to the tune of that amount.

29 21 When Plainridge opens, prior to any full casinos in Massachusetts, I estimate that such imports of gaming services will drop to just $719 million (Exhibit 9-1). At the same time, the residents of other states (primarily Rhode Island) are projected to spend $25.7 million at Plainridge. Net imports of gaming services would then amount to just $693.5 million, an improvement of $62.3 million from the situation without Plainridge. The corresponding calculations for my reference two-full-casino case (Suffolk Downs and Springfield) are presented in Exhibits 9-2A (if Plainridge were not open) and Exhibit 9-2B, in which I assume it is. Under this scenario, Plainridge is projected to be responsible for a $54.2 million improvement in net exports of gaming services (less than when it is the only game in town due to competition from the full casinos). Finally, the corresponding calculations for my Full Competition scenario (casinos at Suffolk, Springfield, and Taunton) are presented similarly in Exhibits 9-3A and 9-3B. Under this scenario, Plainridge is projected to improve Massachusetts s balance of trade in gaming services by $36.2 million (reduced again due to the greater competition from the additional full casino). 16 Impacts on Consumer Spending Per Capita In order to assist Dr. Jeffrey Marotta with his assessment of the likely impacts of the new slot facility on rates of problem gambling among the residents of Plainville, I have used my gravity 16 The curious reader may wonder why my projections for the amounts that the residents of other states will spend at Plainridge actually rise from Exhibit 9-1 to 9-2B (and in part, to 9-3B). This is because with greater competition, Plainridge will be less crowded. When it is the only game in Massachusetts, it will be packed on Friday and Saturday nights. When it becomes less packed, slightly more (though in absolute terms, still a very small number) of the residents of Rhode Island and Connecticut will venture to give it a try. This effect occurs in Massachusetts, too, but is swamped by the direct diversion of Massachusetts customers to the new casinos.

30 22 model to develop estimates for the average amounts that the residents of Plainville (and neighboring communities) will spend on slot-machine gaming under the various scenarios on a per capita basis. These figures are presented in Exhibit 10. When Plainridge opens, electronic gaming machines will become more accessible not only to the residents of Plainville, but also to the rest of Massachusetts, though the effect will be most substantial in Plainville itself. With increased accessibility, the evidence is clear that rates of problem gambling rise, and therefore the associated social impacts. Dr. Marotta uses my estimates regarding spending per adult to develop his projections for such impacts; for the details, see his reports. I would observe here that while the resulting impacts are correlated with spending per adult, the proportion is not strictly one-to-one. In particular, these projections do not imply that gamblingrelated problems will rise at the same rate as consumer spending. The explanation for this is simply that casino gambling is not the only cause of problem gambling, nor more properly the only vehicle that problem gamblers may use. There exist many other such vehicles whether casinos are in town or not including traditional lottery games, scratch tickets, pari-mutuel wagering, social card games, sports betting, and, increasingly, gambling on the internet. The data show that accessibility to electronic gaming devices is indeed a very significant factor in the development and progression of gambling problems, but it is far from the only factor.

31 23 6. Conclusions Exhibit 11 summarizes my projections for the gross gaming revenues of the new slotmachine facility at Plainridge. In this exhibit, these revenues are presented as of stabilized operations and in terms of FY2013 dollars. This exhibit also presents my estimates for slotmachine spending per visitor, and projections for the total number of visitors over each year in thousands. Exhibit 12 presents my projections for gross gaming revenues over five years of operation in terms of then-year dollars, assuming annual escalation due to population growth and rising incomes of 2% per year. I incorporate a 15 percent discount in Year 1 (assumed to be 2015) as the new operation comes up to speed, and a five percent discount in Projected numbers of visitors and average spending per visitor are shown here, too. Over these five years I have assumed that (i) Plainridge is the only gaming facility in Massachusetts in 2015 and 2016; (ii) full casinos open in Springfield and at Suffolk Downs in 2017; and (iii) a casino opens in Taunton, as proposed by the Wampanoag Tribe, in The most representative projections for the Peak Demand and Full Competition scenarios are indicated by shading in light yellow (2016 and 2018, respectively). Exhibit 13 compares my projections for gross gaming revenues with those developed by the Innovation Group for Plainridge Racecourse in November Given the usual range of variation in such projections, we are extremely close.

32 Market Analysis, with Projected Performance of a Slot-Machine Facility at Plainridge Racecourse Exhibits May 28, Jason Street, Arlington, Massachusetts Telephone: Fax: cummingsw@aol.com

33 List of Exhibits Exhibit 1 Gaming Facilities in New England -- Existing and Proposed 2 Detail for Southeastern Massachusetts 3 Recent Performance of Gaming Devices at Tracks and Casinos Nearby 4 Long-Term Trends at Nearest "Mega-Casinos" 5 Long-Term Trends at VLT Facilities in Upstate New York 6 Portion of Model Inputs 7 Gaming-Device "Power Ratings" 8 Projected Slot-Machine Gaming Revenues (series) 9 Total Consumer Spending on Slots & VLT, with Net Exports (series) 10 Projected Annual Slot (and VLT) Spending per Adult 11 Summary of Projections -- FY2013$ 12 Summary of Projections -- Then-Year $ 13 Cummings Projections Compared to Innovation Group's

34 Exhibit 1: Gaming Facilities in New England Existing and Proposed

35 Exhibit 2: Detail for Southeastern Massachusetts

36 Exhibit 3: Recent Performance of Gaming Devices At Casinos and Tracks Nearby (12 mos thru 3/31/13) State Facility Total versus Recent Estimated # Slots/VLTs Slot/VLT Win Prior Win/unit/day "Power 3/31/13 ($000) FY (1) Rating" Nearest (listed by state, not absolute proximity): RI Twin River 4,750 $467, % $ Lincoln RI Newport Grand 1,085 $49, % $ CT Foxwoods 6,128 $556, % $ Ledyard CT Mohegan Sun 5,542 $635, % $ Uncasville NY Empire City 5,300 $545, % (3) $ Yonkers NY Resorts World 5,003 $696,555 na $ Aqueduct / Queens ME Hollywood 950 $52,221 (2) -13.6% (3) $ Bangor ME Oxford 814 $51,794 (2) na $ Representative Comparables: DE Delaware Park 2,293 $166, % (3) $ Wilmington DE Dover Downs 2,452 $164, % (3) $ DE Harrington Raceway 1,818 $90, % (3) $ PA Presque Isle 1,715 $142, % (3) $ Erie PA The Meadows 3,316 $244, % $ near Pittsburgh WV Wheeling Island 1,414 $99,519 (2) -19.7% (3) $ "Power Rating" reflects each facility's ability to attract revenues from the surrounding population based on gravity-model analysis. For discussion, see Appendix A. (1) Many facilities are consolidating numbers of slots. Win/unit/day based on average number of units. (2) Annual rate based on July 2012 (approx. opening, or new competition) - Mar (3) Declines due to opening of new competition.

37

38

39 (first of two pages) Exhibit 6: Portion of Model Inputs Northeast Slot Estimates Travel Times: State County ZIP Code Foxwoods Mohegan Twin R Newport Bangor Oxford Suffolk Everett Plainr Raynhm Worcs Taunton New B MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : MA Worcester MA Worcester MA Worcester MA Worcester MA Worcester Total MA

40 (second of two pages) Exhibit 6: Portion of Model Inputs Northeast Slot Estimates State County ZIP Code MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire MA Berkshire : : : : : : MA Worcester MA Worcester MA Worcester MA Worcester MA Worcester Total MA Impacts: 2012 Adult Milford Palmer Closest Population 200X PCI Dstnce Urban? Prox'y Income Dist-Adj Adults ,010 $28,692 24% 100% 100% 100% 8, ,754 $25,320 20% 100% 100% 96% 1, $24,726 21% 100% 100% 95% ,047 $32,642 24% 100% 100% 100% $34,635 22% 100% 100% 100% ,507 $35,263 21% 100% 100% 100% ,119 $29,182 22% 100% 100% 100% 1, ,659 $33,442 24% 100% 100% 100% 1, ,823 $31,837 21% 100% 100% 100% ,087 $30,719 24% 100% 100% 100% ,217 $36,914 21% 100% 100% 100% ,441 $30,202 30% 100% 100% 100% 1, ,371 $38,004 27% 100% 100% 100% 1, $28,344 21% 100% 100% 100% $34,597 24% 100% 100% 100% ,142 $22,681 19% 100% 100% 92% 2,151 : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : $58,753 54% 100% 101% 100% ,546 $40,461 87% 100% 101% 100% 4, ,031 $43,715 52% 100% 101% 100% 2, ,945 $34, % 100% 101% 100% 21, ,621 $55,952 52% 100% 101% 100% 3,458 4,699,294 1,769,084

41 Exhibit 7: Gaming-Device "Power Ratings" (Total Annual Spending versus benchmark of $720) Northeastern US Medium to Large Markets Rural Markets and Florida Elsewhere Elsewhere Mississippi average Deadwood, SD Cripple Creek, CO (2) Upstate Michigan avg e Laguna Tribe (3 facils), NM San Felipe (ABQ), NM S Dakota Tribes (avg of 8) e Downstream Resort, OK Black Hawk/CC, CO (2) Santa Ana (ABQ), NM Louisiana average Kansas Tribes (avg. of 4) e Salamanca, NY e Other NM (avg. of 9) Diamond Jo Worth, IA Sandia (ABQ), NM Emmetsburg, IA Iowa Tribes (average of 3) e Zia Park (Hobbs), NM Midwest Standard +10% Atlantic City, NJ avg Isleta (ABQ), NM SunRay Park, NM Seneca Niagara (NY) e Upstate Wisconsin avg e Dodge City, KS Tioga Downs, NY IOC Boonville, MO o Vernon Downs, NY Terribles Lakeside. IA (Buffalo) Fairgrounds, NY Harrahs NKCMO Horseshoe / Bluffs Run, IA Taos, NM Grand Falls, IA (S. Falls, SD) Mohegan Sun, CT Riverside, IA Ocean Downs, MD The Downs at ABQ, NM Dubuque Diamond Jo, IA Ameristar Council Bluffs, IA Argosy Sioux City, IA o Mt. Pleasant, MI e Mountaineer Park, WV Dubuque Mystique, IA IOC Marquette, IA o Foxwoods, CT Ameristar KCMO Pocono Downs, PA Argosy Riverside, MO Wisconsin Dells e "Midwest Standard" IOC Waterloo, IA 99.7 Presque Erie, PA 98.8 Michigan City, IN 98.7 Finger Lakes, NY 98.7 Dover Downs, DE 98.0 The Meadows / Pittsburgh 97.4 Ruidoso Downs, NM 97.5 Harrington Raceway, DE 97.0 Harrahs W St Louis 96.9 Saratoga, NY 96.8 Jumers Rock Island, IL 96.2 Wheeling, WV 95.8 Detroit (avg / 3 facils) 96.0 o St Jo MO 95.2 o Penn National / Harrisburg, PA 94.7 Harrahs Council Bluffs, IA 94.6 Mount Airy / Pocono, PA 94.4 Charles Town, WV 93.5 Belterra, Florence, IN 93.7 o Mark Twain, MO 93.7 o Delaware Park 93.3 Prairie Meadows, IA 93.2 Clinton, IA 93.1 Ameristar St Chas, MO 93.1 Harrahs Joliet, IL 92.8 o Metropolis, IL/KY 92.7 o Catfish Bend Burlington, IA 92.3 East St Louis, IL (2 facils) 92.0 o Batavia, NY 91.0 Midwest Standard -10% Chester, PA 89.1 Ameristar, E Chicago IN 88.9 o Green Bay, WI 88.3 e Monticello, NY 88.5 St. Louis, MO (2 facils) 88.3 Sands Bethlehem, PA 87.9 IOC Bettendorf, IA 88.2 o Twin Lincoln, RI 87.8 Indiana Live 87.3 Caruthersville, MO 87.3 o Hammond, IN 87.2 The Rivers / Pittsburgh 86.0 IOC KCMO 85.8 o Rising Sun, IN 85.8 o Hoosier Park, IN 85.5 Hollywood, Lawr'burg, IN 85.4 o Rhythm City, IA 85.4 o Parx / Philadelphia 84.4 French Lick, IN 83.8 Sugarhouse / Philadelphia 82.6 Majestic Star, Gary IN 82.4 o Elgin (Chicago) IL 81.7 o Louisville, KY/IN 81.1 o Midwest Standard -20% Milwaukee, WI 79.9 e o Aurora (Chicago), IL 79.7 o Newport Grand, RI 79.0 Joliet Empress, IL 78.5 o Pompano Park, FL 77.1 Resorts Aqueduct, NY 77.1 o Bangor, ME 76.2 Oxford, ME 75.9 Gulfstream Park, FL 75.4 Empire Yonkers, NY 74.5 o Mardi Gras / Hollywood, FL 73.1 Peoria, IL 72.3 o Arundel Mills, MD 71.9 KCKS 7th St Casino 71.8 e Mardi Gras, WV 71.6 (Class II slots) Cecil County, MD 71.3 Evansville, IN 69.2 o Sunland Park, NM 69.0 o Montana VLTs 69.0 Magic City / Flagler, Miami, FL 68.1 Calder Race Course, Miami, FL 65.0 South Dakota VLTs 64.6 Miami Jai-Alai & Casino 57.5 a a = annual rate, e = estimated, o = old boat or capacity-constrained market

42 Full Casinos: Exhibit 8-0: Projected Total Slot & VLT Gaming Revenues (in FY2013$) Null Case Projected: Assumed # Slots Power $ / slot / day Slot Win Table Win Total Win Rating ($mn) (na) ($mn) Suffolk Downs Everett Milford Taunton New Bedford Springfield Palmer Subtotal 0 $0.0 $0.0 $0.0 Slot Parlor: Plainridge Raynham Other Total MA 0 $0.0 $0.0 $0.0 All scenarios assume: No new on / off-ramps, but otherwise good micro-access at all facilities. Existing legislation in Mass & competition from other states (plus tables at Twin River).

43 Full Casinos: Exhibit 8-1: Projected Total Slot & VLT Gaming Revenues (in FY2013$) Peak Demand - Plainridge with No Other Casinos in Mass. Projected: Assumed # Slots Power $ / slot / day Slot Win Table Win Total Win Rating ($mn) (na) ($mn) Suffolk Downs Everett Milford Taunton New Bedford Springfield Palmer Subtotal 0 $0.0 $0.0 $0.0 Slot Parlor: Plainridge 1, * $430 $196.4 $196.4 Raynham Other Total MA 1,250 $196.4 $0.0 $196.4 * Power Rating below benchmark due to crowding. All scenarios assume: No new on / off-ramps, but otherwise good micro-access at all facilities. Existing legislation in Mass & competition from other states (plus tables at Twin River).

44 Full Casinos: Exhibit 8-2: Projected Total Slot & VLT Gaming Revenues (in FY2013$) Plainridge with Two Casinos Projected: Assumed # Slots Power $ / slot / day Slot Win Table Win Total Win Rating ($mn) (na) ($mn) Suffolk Downs 4, * $389 $568.3 $0.0 $568.3 Everett Milford Taunton New Bedford Springfield 4, * $343 $501.5 $0.0 $501.5 Palmer Subtotal 8,000 $1,069.8 $0.0 $1,069.8 Slot Parlor: Plainridge 1, * $342 $156.1 $156.1 Raynham Other Total MA 9,250 $1,225.9 $0.0 $1,225.9 * Power Rating below benchmark due to crowding. All scenarios assume: No new on / off-ramps, but otherwise good micro-access at all facilities. Existing legislation in Mass & competition from other states (plus tables at Twin River).

45 Full Casinos: Exhibit 8-3: Projected Total Slot & VLT Gaming Revenues (in FY2013$) Plainridge with Three Casinos ("Full Competition") Projected: Assumed # Slots Power $ / slot / day Slot Win Table Win Total Win Rating ($mn) (na) ($mn) Suffolk Downs 4, * $353 $514.8 $0.0 $514.8 Everett Milford Taunton 4, * $310 $452.3 $0.0 $452.3 New Bedford Springfield 4, * $334 $488.3 $0.0 $488.3 Palmer Subtotal 12,000 $1,455.4 $0.0 $1,455.4 Slot Parlor: Plainridge 1, $285 $129.8 $129.8 Raynham Other Total MA 13,250 $1,585.3 $0.0 $1,585.3 * Power Rating below benchmark due to crowding. All scenarios assume: No new on / off-ramps, but otherwise good micro-access at all facilities. Existing legislation in Mass & competition from other states (plus tables at Twin River).

46 Exhibit 9-0: Total Consumer Spending on Slots & VLTs (FY2013 $mn) Null Case TO: Other Subtotal Other Total Plainridge Mass. AT casinos States' FROM FROM: Casinos in Mass. Casinos Mass. Massachusetts Residents $0.0 $0.0 $0.0 $755.8 $755.8 Residents of Other States Connecticut $0.0 $0.0 $0.0 New York Northern New England Rhode Island Further Afield Subtotal FROM Other States $0.0 $0.0 $0.0 Grand Total AT Mass. Casinos $0.0 $0.0 NET EXPORTS: -$755.8 (difference) Memo: Norfolk County Net Exports -$89.4

47 Exhibit 9-1: Total Consumer Spending on Slots & VLTs (FY2013 $mn) Peak Demand - Plainridge with No Other Casinos in Mass. TO: Other Subtotal Other Total Plainridge Mass. AT casinos States' FROM FROM: Casinos in Mass. Casinos Mass. Massachusetts Residents $170.7 $0.0 $170.7 $719.1 $889.9 Residents of Other States Connecticut $2.7 $0.0 $2.7 New York Northern New England Rhode Island Further Afield Subtotal FROM Other States $25.7 $0.0 $25.7 Grand Total AT Mass. Casinos $196.4 $0.0 $196.4 NET EXPORTS: -$693.5 (difference) NET CHANGE vs. Null Case $62.3 Memo: Norfolk County Net Exports $73.5

48 Exhibit 9-2A: Total Consumer Spending on Slots & VLTs (FY2013 $mn) Two-Casino Case (w/o Plainridge) TO: Other Subtotal Other Total Plainridge Mass. AT casinos States' FROM FROM: Casinos in Mass. Casinos Mass. Massachusetts Residents $0.0 $745.5 $745.5 $390.8 $1,136.3 Residents of Other States Connecticut $0.0 $156.3 $156.3 New York Northern New England Rhode Island Further Afield Subtotal FROM Other States $0.0 $358.8 $ Grand Total AT Mass. Casinos $1,104.3 $1,104.3 NET EXPORTS: -$32.0 (difference) NET CHANGE vs. Null Case $723.7 Memo: Norfolk County Net Exports -$108.2

49 Exhibit 9-2B: Total Consumer Spending on Slots & VLTs (FY2013 $mn) Plainridge with Two Casinos TO: Other Subtotal Other Total Plainridge Mass. AT casinos States' FROM FROM: Casinos in Mass. Casinos Mass. Massachusetts Residents $127.8 $716.4 $844.2 $359.6 $1,203.8 Residents of Other States Connecticut $2.8 $155.4 $158.2 New York Northern New England Rhode Island Further Afield Subtotal FROM Other States $28.3 $353.4 $ Grand Total AT Mass. Casinos $156.1 $1,069.8 $1,225.9 NET EXPORTS: $22.1 (difference) NET CHANGE vs. Null Case $777.9 vs. No Casino at Plainridge: $54.2 Memo: Norfolk County Net Exports $24.0

50 Exhibit 9-3A: Total Consumer Spending on Slots & VLTs (FY2013 $mn) Three-Casino Case (w/o Plainridge) TO: Other Subtotal Other Total Plainridge Mass. AT casinos States' FROM FROM: Casinos in Mass. Casinos Mass. Massachusetts Residents $0.0 $1,027.9 $1,027.9 $266.0 $1,293.9 Residents of Other States Connecticut $0.0 $170.9 $170.9 New York Northern New England Rhode Island Further Afield Subtotal FROM Other States $0.0 $479.9 $ Grand Total AT Mass. Casinos $1,507.8 $1,507.8 NET EXPORTS: $213.9 (difference) NET CHANGE vs. Null Case $969.7 Memo: Norfolk County Net Exports -$124.0

51 Exhibit 9-3B: Total Consumer Spending on Slots & VLTs (FY2013 $mn) Plainridge with Three Casinos ("Full Competition") TO: Other Subtotal Other Total Plainridge Mass. AT casinos States' FROM FROM: Casinos in Mass. Casinos Mass. Massachusetts Residents $101.3 $984.3 $1,085.6 $249.6 $1,335.1 Residents of Other States Connecticut $3.2 $169.7 $172.9 New York Northern New England Rhode Island Further Afield Subtotal FROM Other States $28.6 $471.1 $ Grand Total AT Mass. Casinos $129.8 $1,455.4 $1,585.3 NET EXPORTS: $250.1 (difference) NET CHANGE vs. Null Case $1,005.9 vs. No Casino at Plainridge: $36.2 Memo: Norfolk County Net Exports -$13.9

52 Exhibit 10: Projected Annual Slot (and VLT) Spending per Adult Scenario: Plainville Wrentham Foxborough Mansfield N Attleboro Current (FY2013) (1) $249 $217 $239 $237 $ Null Case (Tables at Lincoln) $254 $221 $243 $242 $ Plainridge with No Casinos $588 $417 $388 $405 $543 (2) 2A. Two Casinos w/o Plainridge $254 $221 $244 $242 $304 2B. Plainridge w Two Casinos $624 $438 $404 $424 $571 (2) 3A. Three Casinos w/o Plainr. $279 $257 $290 $324 $332 3B. Plainridge with 3 Casinos $667 $470 $435 $464 $615 (2) Compare vs. Benchmark $720 Current (FY2013) 35% 30% 33% 33% 41% 0. Null Case (Tables at Lincoln) 35% 31% 34% 34% 42% 1. Plainridge with No Casinos 82% 58% 54% 56% 75% (2) 2A. Two Casinos w/o Plainridge 35% 31% 34% 34% 42% 2B. Plainridge w Two Casinos 87% 61% 56% 59% 79% (2) 3A. Three Casinos w/o Plainr. 39% 36% 40% 45% 46% 3B. Plainridge with 3 Casinos 93% 65% 60% 64% 85% (2) Increment versus Appropriate "Pre-Plainridge" Case - as Percentage of $ Plainridge with No Casinos 46% 27% 20% 23% 33% 2B. Plainridge w Two Casinos 51% 30% 22% 25% 37% 3B. Plainridge with 3 Casinos 54% 30% 20% 19% 39% : i.e., Plainridge will add about 50% of the "burden" of a "full casino" in town Dollar Increment versus Current Day $ / Adult 1. Plainridge with No Casinos 134% 90% 61% 69% 80% 2B. Plainridge w Two Casinos 149% 100% 67% 77% 90% 3B. Plainridge with 3 Casinos 156% 98% 61% 59% 95% : i.e., increment due to Plainridge is about 150% of current spending / adult (1) Current spending by Plainville residents is approximately 157% x Mass. State average of $159. (2) Plainville spending rates are projected to be less than my benchmark $720 (100%) because (a) no tables, (b) modest rates of spending on player rewards, (c) slightly more than 7 minutes average travel time, and (d) in Scenarios 1 and 2B, crowding (weekend evenings in particular unable to accomodate peak demand).

53 Exhibit 11: Projected Total Gaming Revenues For a Slot-Machine Gaming Facility at Plainridge (FY2013$) Plainridge "Peak Demand" ("as if" in FY2013) Slot Win ($/day) Impact ($/vis) # Visitors (Cumulative) (000) (No New Gaming Elsewhere) $196.4 mn ($431) $84.5 2,323 (except tables at Twin River) Two Mass. Casinos open: Springfield and... Region A: Suffolk Downs (4,000 slots) $156.1 mn ($342) -21% $80.1 1,948 Everett (4,000 slots) $153.4 mn ($336) -22% $79.8 1,922 Milford (4,000 slots) $150.3 mn ($330) -23% $79.5 1,891 Region C: Taunton (4,000 slots) $151.1 mn ($331) -23% $79.6 1,899 New Bedford (4,000 slots) $167.4 mn ($367) -15% $81.3 2,058 Three Mass. Casinos open: "Light" competition (Suffolk, New Bedford & Sprgfield) $144.6 mn ($317) -26% $78.9 1,834 "Full" competition (Suffolk, Taunton & Springfield) $129.8 mn ($285) -34% $77.2 1,681 "Intense" competition (Milford, Taunton & Palmer) $115.5 mn ($253) -41% $75.7 1,526 Notes: All scenarios assume: 51% retention rate, therefore reasonable (though modest) CapEx and player rewards. Good micro-access. 24/7 operation. 1,250 machines, including electronic table games. Slot "Power Rating" (i.e., "everything else") = 95(-), similar to comparable facilities. All projections are as of "stabilized operations," and assume standard patterns of investment in facilities, marketing and player rewards. These may not be optimum here. Table games at Twin River and Newport. No gaming in NH.

54 Exhibit 12: Projected Total Gaming Revenues For a Slot-Machine Gaming Facility at Plainridge (Then-Year $) Projections in FY2013$: Slot Win ($/day) ($/vis) # Visitors (000) (No New Gaming Elsewhere) $196.4 mn ($431) $84.5 2,323 (except tables at Twin River) Springfield + Suffolk Downs $156.1 mn ($342) $80.1 1,948 (assumed as of 1/1/2017) Springfield + Suffolk + Taunton $129.8 mn ($285) $77.2 1,681 (assumed as of 1/1/2018) Projections in Then-Year Dollars: (2.0%) Slot Win ($/day) ($/vis) # Visitors Year Base $ Escal'n Start-Up (000) 2015 $ $175.4 mn ($385) $85.4 2, $ $200.0 mn ($438) $89.4 2, $ $170.6 mn ($374) $87.6 1, $ $144.8 mn ($317) $86.1 1, $ $147.7 mn ($324) $87.8 1,681 Notes: All scenarios assume: 51% retention rate, therefore reasonable (though modest) CapEx and player rewards. Good micro-access. 24/7 operation. 1,250 machines, including electronic table games. Slot "Power Rating" (i.e., "everything else") = 95(-), similar to comparable facilities. All projections are as of "stabilized operations," and assume standard patterns of investment in facilities, marketing and player rewards. These may not be optimum here. Table games at Twin River and Newport. No gaming in NH.

55

56 Market Analysis, with Projected Performance of a Slot-Machine Facility at Plainridge Racecourse Appendix: Details of the Gravity-Model Methodology May 28, Jason Street, Arlington, Massachusetts Telephone: Fax: cummingsw@aol.com

57 Appendix Details of the Gravity-Model Methodology My projections for the likely performance of new gaming facilities are based upon analyses of the experience of the most comparable operations elsewhere in the region and more broadly all across the United States. I use gravity models as a key element of this process. This methodology has been refined over the years as others and I have applied it to assessing the performance of many gaming facilities, both existing and proposed. It is based essentially on the demographics of the areas surrounding each facility, in particular the number of adults residing at various distances, and the ratio of actual revenues obtained to such adult populations at existing facilities. Access time, not mileage per se, and population density are the most critical variables. (A bibliography is attached.) To illustrate the relationships among casino revenues, population, and distance, Exhibit A-1 presents a graph which compares rates of visitation versus distance for the casinos of Mississippi, based upon statewide survey data. There is clearly a relationship between patronage and distance: the greater the distance the customer has to travel, the lower the number of visits. Fewer patrons are willing to travel longer distances, and when they do, they usually visit less often. (Offsetting this slightly, when they do visit, they typically spend more on each occasion than nearby customers who visit more frequently distance acts as a filter to deter more casual fans.) In addition, the further you live from these casinos, the closer you generally get to competing casinos in other states, further reducing your rate of visiting Mississippi. Because rates of visitation appear to decline so dramatically as distance increases, and because the scale is so large when looking at statewide data such as these from Mississippi, it is useful to transform this data by taking logarithms ( log-transforming the data, as economists say). Exhibit A-2

58 2 presents the Mississippi data in such fashion, and, to my eyes, at least, presents a pattern that comes across more clearly. When we exclude the most distant data (beyond 250 miles, where competition, rather than distance, usually becomes the most critical factor), regression analysis indicates a relationship that is indeed fairly robust (Exhibit A-3). I have analyzed such data from a wide variety of markets, and have estimated that in general, over a reasonable range of distances the aggregate elasticity of spending with respect to distance is roughly -0.7, that is, consumers total spending declines in somewhat less than direct proportion to the distance to be traveled. 1 When, however, several facilities compete within the same (or closely connected) market(s), the customer overwhelmingly prefers the closest. It appears that in this respect slot machines (and similar video lottery terminals, or VLTs) behave in a fashion very similar to many other retail markets, in which the relative attraction of each outlet is roughly inversely proportional to the distance squared. 2 Using these parameters to account for the relationships with distance and demographic data for each county in gaming markets across the United States (and in cases such as Massachusetts for each zip code), I have calculated the distance-adjusted adult population surrounding each slot or VLT facility or close group of such facilities in each market. (A portion of my model for the Northeastern 1 This is a relatively long-distance attraction; if you double the distance, revenues decline by about 38%. For comparison, pari-mutuel betting at race tracks generally exhibits a distance coefficient of about - 1 to -1.2: if you double the distance, visitation declines by 50% or more. Generically, this type of relationship is called a gravity model, because it is similar to Newton s law of gravitation (for which the distance factor would be -2.0: if you double the distance, the attraction declines by a factor of 2 2, or four). With respect to travel time, the elasticity appears to be slightly less; I estimate A relationship sometimes called Reilly s Law of Retail Gravitation, based upon its mathematical identity with Newton s Law, above. David Huff and others have extended these models further with many retail applications, so they are known more generally today as Huff models.

59 3 U.S. is presented in Exhibit A-4, which extends over two pages.) Dividing the total revenues, or spending, in each existing market by these population figures results in ratios measuring revenues, or consumer spending, per distance-adjusted adult. The gravity model, based upon Reilly s Law noted above, then distributes these adults (and so, by proxy, their spending) across the different gaming facilities, or closely-situated groups of gaming facilities, to which they have access. By summing across geographic areas, we can then estimate the sources of revenues (again, consumer spending) for each such group of facilities. These models can also be used to compare different markets and facilities against one another. Statistics of this type are presented for the major regional gaming markets of the U.S. in Exhibit A-5, listed in order of estimated slot (or VLT) spending per person. 3 Again, the gravity-model procedure simply puts the different markets onto a common footing in terms of performance, abstracting out differences due to the varying distributions of population around each facility. The figure for each market represents the amount that the average adult that lives within ten minutes of (legal) gaming devices spends on them each year. 4 Note that this exhibit extends over two pages. As benchmarks, I have inserted several horizontal yellow bars, which represent what I call Midwest Standard performance ($720 per distance-adjusted adult per year), plus ten percent, minus ten percent, and minus twenty percent. In the Midwest, modern casino facilities in populous markets that are more or less typical consistently cluster A point of terminology: Huff describes the general decline with distance (as opposed to the competitive decline) as friction. I think this is a very useful way to look at this process, particularly with respect to the traffic-congested markets of the Northeast as opposed to the more rural Midwest. 3 These figures do not include relatively modest amounts spent at casinos in Las Vegas, the Caribbean, and other remote destination resorts in the U.S. and abroad. 4 There is nothing special about the ten-minute figure; it is simply a benchmark to represent convenient access.

60 4 around the $720 benchmark. In other parts of the country, as indicated in the columns to the left and to the right, the dispersion is somewhat greater. In an attempt to simplify comparisons among markets, and to clarify discussion of the principles involved, I have converted these dollar figures into what I call power ratings in Exhibit A- 6 (pardon the small print, but I sometimes think it useful to see all these data points on one page.) Midwest Standard spending of $720 per year translates into a power rating of 100; ten percent higher ($770) translates into 110, and ten percent less into 90. I think these ratings are intuitively more comprehensible than large dollar figures that have not been normalized versus some standard for comparison. It may be helpful to consider these power ratings as a kind of extension of the fair share concept that is often used to compare different facilities in the gaming industry. If, for example, all the slot machines in a given market average $200 in win per day as a group, a facility at which the machines win $240 per day is said to do 120% of its fair share. One that does $180/day/machine, on the other hand, wins just 90% of its fair share. My extension to power ratings adds analysis of the surrounding demographics. If there are many people and few machines (like the Chicago area, for example), high wins per machine per day should be expected. Harrah s Joliet, as an extreme example, wins roughly $500/machine/ day. In Iowa, on the other hand, there are many more machines and many fewer people. The newly landbased Wild Rose casino at Clinton, for example, wins just $177/machine/day. Based on my gravitymodel analysis, however, I estimate the power rating for the Clinton casino at 93.1, while that for Harrah s Joliet is just Despite its much lower win/slot/day, the Wild Rose at Clinton actually does better in drawing from the population which surrounds it that population is simply much smaller at Clinton than at Joliet. And some other casinos in Iowa actually do even better, such as the casino at

61 5 Riverside, Iowa, the Dubuque Diamond Jo, and the Mystique (formerly Dubuque Greyhound Park). My analyses indicate that these casinos have slot power ratings ranging from 101 to 103, while their wins/slot/day are in the same ballpark as the Wild Rose ($150-$180/day). 5 When analyzed in this way, the range of experience across the diverse spectrum of markets depicted in Exhibits A-5 and A-6 is, in my opinion, not all that wide. The difference between the best markets (Mississippi, South Dakota, Colorado, New Mexico and Louisiana) and the worst (several highly urban casinos in New York and South Florida) amounts to roughly a factor of two. Most of these markets fall into the range of $600 to $800 in annual spending per distance-adjusted adult or in terms of power ratings, from 80 to 110. As indicated by the columns in these exhibits, I have divided the broad universe of markets into three groups: the Northeastern U.S. and Florida in the first column, medium to large markets elsewhere in the second column, and very rural markets in the third column. Rural markets tend to do better than others for three reasons: (i) it is easier to get around rural areas than urban ones (the friction is less a twenty-minute drive on a rural highway is generally far less challenging than one of similar duration in urban or even suburban traffic), (ii) there is less competition from other commercial entertainment activities, and (iii) there is likely some survival bias in the data rural facilities often 5 To press my point further, the Horseshoe Casino at the Bluffs Run greyhound track has a power rating that is very similar to these (105.6), but because it serves a more densely-populated market (Omaha), its win per slot per day is significantly higher ($247). I should perhaps explain at this point that because of all the ways in which I use power ratings in my models, the difference between 93 and 105 (for example) ultimately results in much more than a twelve percent difference in performance. I use the power ratings to modulate (i) average spending per person in the market, (ii) market share for each facility, and (iii) the reach of each facility at greater distances (the distance coefficient that represents the competitive interactions of the Reilly and Huff models). Other things being equal (i.e., the surrounding demographics), one point of power rating typically translates into 3-4% change in performance.

62 6 serve such small markets that only the best survive. Thus, most of the best facilities in the top right corners of Exhibits A-5 and A-6 are very country. 6 More competitive markets also appear to attract higher rates of spending. This is true even for the Northeastern U.S. and for rural areas, but I have placed all of the markets in each area into their respective geographic zones for ease of comparison. Outside of the most rural markets, those in Mississippi demonstrate the best performance. Tunica is, of course, somewhat rural, but the more urban casinos on the Gulf Coast and at Vicksburg seem to do nearly as well. In my view, this is because there are no statutory limits on the number or size of casinos in Mississippi, its tax rate is very low, and there are multiple properties at most locations, so all of its markets are highly competitive. These casinos therefore attract high rates of spending. Similar factors apply to New Mexico, both in rural areas and in the metropolitan area of Albuquerque. And while their tax rates are somewhat higher, the Colorado casinos, while restricted to three remote former mining towns (and until recently to $5 bets), are also highly competitive, as are most of the major (and minor) markets of Louisiana and Iowa, and many of the rural markets in other states in the top right corner. As we move down the middle column, we generally find less competitive conditions, with areas such as Chicago, Detroit, and Milwaukee, where the number of facilities and/or gaming devices is nowhere near sufficient to meet the demand for them, and/or the markets are constrained by cramped conditions, on riverboats or ashore. As a result of these capacity-constrained conditions, spending per 6 Even in the left-hand column, the data points at the top are generally very rural, or else large but remote destination casinos, while casinos in the most urban settings (The Rivers, Parx, and Sugarhouse in Pennsylvania, Resorts World and Empire City in New York, and three casinos in Miami) fall near the bottom.

63 7 (distance-adjusted) adult is relatively low in these areas. (Conversely, spending per machine is typically [but not always] very high, as people are figuratively lined up at the machines to play them.) 7 Even in these markets, however, slot spending per distance-adjusted adult generally ranges from $550 to $650 per year, not all that far below the $720± that most competitive casino markets demonstrate and even some less-competitive markets achieve. At the bottom of the left-hand column are some of the VLT facilities in New York State, Rhode Island, and West Virginia, along with most of the slot-machine facilities in South Florida. The facts that these tend to involve video lottery terminals, and are located at race tracks, are in my opinion of little import. In most cases (aside from New York), these VLTs are identical to the slot machines found in casinos elsewhere. 8 It is, however, surely no coincidence that these four jurisdictions have some of the highest tax rates on gaming devices found anywhere in the U.S. With high tax rates, only modest investments in new and improved facilities can earn a reasonable return. As a result, the facilities in New York were initially very modest indeed, and, with a few exceptions, most of those in the other states as well. High tax rates also limit the operators ability to spend effectively on promoting their gaming product, including in particular player rewards programs. In highly competitive jurisdictions such as Iowa and New Jersey, casinos spend more than twenty percent of their gaming revenues on such promotion. With less than fifty percent of the gross retained by the gaming facilities in the lower left corner, spending any significant fraction of that amount is impossible. 7 Markets can effectively be capacity-constrained even when, as at the Rising Sun and Belterra casinos in Indiana southwest of Cincinnati today, and at some of the New York and Rhode Island VLT facilities, win/machine/day is not at astronomical levels. If the major issues are accessibility and attractiveness (simply in terms of spaciousness, amenities, and/or quality of machines, not necessarily glitz ), players may indeed not be lined up to play as they are in other jurisdictions where the unsatisfied demand is far more obvious. 8 The gaming facilities at race tracks in Pennsylvania, Delaware and West Virginia, now offer table games as well, so they are now truly full (-spectrum) casinos.

64 8 Florida, with many facilities at the bottom of the left-hand column, initially followed a similar model, with a tax rate of 50% on slot gaming at the pari-mutuel facilities in Miami-Dade and Broward Counties that it authorized in With very modest investments at most of the facilities and little to spend on player rewards, the slots at the South Florida tracks have so far performed in a fashion very similar to the worst of those in the Northeast. Their tax rate was reduced to 35% in 2009, but because they were designed in much leaner times, their performance (like their facilities) still tends to lag their peers elsewhere. 9 In today s competitive environment, attractive facilities and intensive promotion are essential to obtaining high volumes of revenue. In the 1990s, when race tracks first installed slot machines and VLTs, it was often sufficient to put slots in a barn and attract large numbers of customers. That is not the case today. If facilities do not meet competitive standards of attractiveness and marketing, they will see many fewer customers than those that do. The data do in fact demonstrate a strong relationship between tax rate, or more precisely its converse, the retention rate that casinos are allowed to keep, 10 and their ability to generate revenues 9 In addition to gaming facilities and player rewards (initially) designed on a shoestring, the slot operations at the pari-mutuel facilities in South Florida suffer from serious traffic congestion and access issues, and face substantial competition from first-class gaming facilities operated by the Seminole Tribe immediately nearby. Moreover, smoking is allowed at the Seminole facilities, but not (indoors) at the pari-mutuels. Still, four out of the five South Florida race track facilities then operating showed double-digit growth in over the past two years, with three in the range of 18-19%. 10 In addition to taxes on gross revenues, the retention rate also reflects the subtraction of mandatory purse payments to horsemen, breeders funds, and other social mandates (in Iowa, for example, the gaming license must technically be held by a public-benefit non-profit entity, which typically receives about 4% of GGR.) In Delaware, New York and West Virginia, retention rates vary by facility. The figures shown for each state are arithmetic averages of those for each gaming facility (i.e., they are not weighted by GGR). I have excluded Florida, Indiana and New Mexico from this analysis: Indiana because its race track facilities are handicapped by amortization of enormous up-front license fees one has just emerged from Chapter 11, and one is still going through it. (Those tracks retain roughly 49%, and their performance to date has been in the mid-eighties.) I have omitted New Mexico (retention rate 53.8%, average power rating

65 9 as measured by power rating. These data are graphed in Exhibit A-7, and aggregated by state in the table below: State Retention Rate Average Power Rating (FY2013) (FY2013) Iowa % New York % 94.0 Delaware % 96.1 Pennsylvania 45.0% 91.3 West Virginia 44.0% 90.5 Maryland % 84.5 Rhode Island % 83.4 Note that my commentary here has tended to focus on slots at tracks. Such facilities face the highest tax rates on gaming in the U.S., as well as deductions for purses and the like, so this effect is most obvious at race track casinos. It is clear, however, that this phenomenon applies to standalone 109.1), because three of its five race tracks are located in rural areas that are very remote, which boosts their ratings substantially. I have excluded Florida because (a) the tax rate was reduced there so recently, and (b) I do not have precise figures regarding purse contributions, and therefore effective retention rates, at its track slot facilities. With retention rates (formerly) in the 40s and power ratings in the 70s, however, its facilities would generally fall somewhat below the curve set by the others. 11 The Iowa data are for the three casinos at its race tracks only. Its non-race-track casinos retain approximately 76%. Several of these are, however, highly rural (while conversely, four remain old-style riverboats), so I have limited my analysis here to the markets in Iowa that are most similar to those in the East. Adding other Mid- and Western states to this analysis would pose similar issues. 12 For comparability among the different states, the retention rates presented here assume that the operators pay for the gaming machines. In Delaware, New York, Maryland and Rhode Island, these are actually provided by the State Lottery, but in the other states must be provided by the tracks. The retention rates shown in my charts and this table therefore include six percent to represent machine costs in those circumstances in which they are actually paid by the state.

66 10 casinos as well, i.e., those not attached to race tracks, and I have included the relevant data for them in each state in the analysis described above. With an effective state tax rate of only 49% (including 9% for the racing industry), the retention rate at Plainridge will be 51%. This will be slightly lower than that at most facilities in New York State, but Plainridge will have a broader (and more attractive) range of slot machines than New York s VLTs. The retention rate at Plainridge will be slightly better than those at the corresponding facilities in Pennsylvania, Delaware and West Virginia. It will be much better than the retention rate at its nearest competitor Twin River, which keeps less than one-third of its VLT win. 13 I therefore believe that the new slot facility at Plainridge Racecourse will demonstrate a power rating on the order of 95, similar to those at the most comparable facilities in New York, Pennsylvania, and Delaware, rather than the lower levels of performance found in Rhode Island (and Maryland). This corresponds to average annual spending of $684 per distance-adjusted adult. Projections for new facilities based on similar gravity models have proved reasonably accurate in the past, in my opinion, and more often conservative than aggressive. Exhibit A-8 presents a tabulation of actual results versus my projections for facilities that have actually been built over the past ten years. 13 It actually keeps less than 30%, but as I described previously, Twin River s VLTs are provided by the State, while Plainridge will have to buy (or lease) its own. I am trying to keep the comparison as apples-toapples as possible.

67 11 Projections for Plainridge For the new slot-machine facility in Plainville, I have specifically assumed that: o The gaming facility will be comparable to those of existing race track casinos in the Northeast in terms of access, appearance, spaciousness and amenities. I have assumed that micro-access with respect to ingress and egress will be good. I have assumed the parking structure that is now almost complete, but no hotel on site. o I have examined multiple scenarios regarding competition: 1. In the expectation that Massachusetts s slot facility will be licensed, built and put into operation well before any of its full-scale casinos, a scenario (which I would expect to extend for one to two years) in which Plainridge would be the only new gaming facility in Southern New England; and then 2. A range of scenarios in which one to three full casinos, on scales close to those of Foxwoods and Mohegan Sun, are put into operation at Suffolk Downs, Everett, Milford, Taunton, New Bedford, Palmer, and/or Springfield, Massachusetts. I have assumed 4,000 slots at each of these facilities, compared to 6,000± currently at each of the casinos in Connecticut, and 4,750 VLTs at Twin River in Rhode Island. 14 o The performance of the Plainridge facility and the underlying propensity to spend of the population surrounding it will be similar to those of existing facilities in comparable situations. I have specifically assumed average annual spending of $684 per distanceadjusted adult, which corresponds to a power rating of 95. o This assumption applies to a time of stabilized operations, which is typically one to three years down the road from the opening of a new gaming facility, and assumes industrystandard patterns of investment in bricks and mortar and in player rewards. Given that a substantial change in the competitive landscape can be foreseen, these patterns (and in fact the entire concept of stabilized operations ) may not be appropriate here. o The existing casinos of Connecticut continue to operate largely as they do today, while the VLT facilities in Rhode Island add table games as planned. o No other new gaming facilities are developed in Southern New England or nearby New York.

68 12 To develop projections based on these assumptions, I took the detailed model illustrated (in part) in Exhibit A-4, calculated the numbers of distance-adjusted adults likely to patronize the new facilities, and applied the appropriate average rates of spending to each. The results are described in the main body of this report. All my analyses and projections are based on the performance of facilities elsewhere in what I refer to as Fiscal Year 2013 (the twelve months ended March 31, 2013), and are therefore calculated initially in terms of FY2013 dollars. I then extrapolate to future years assuming normal growth, due to rising local population, incomes, and inflation, at 2% per year. As a new gaming facility works out its kinks, however, there is typically an initial transient of five to 15 percent in the first year or two. I have assumed the first year here will likely be at the steep end of this range (-15%) because the new facility will have to develop its players list and rewards programs in the face of what will likely be strenuous efforts by its existing competitors to retain their current players. As described above, however, the situation at Plainridge will be complicated by the substantial dip that will likely occur when the full casino(s) open in Massachusetts. Industry-standard patterns of investment may therefore not be optimal here, and the resulting trajectory toward stabilized operations may thus differ from those typically seen elsewhere. 14 With modern multi-denomination (and in many cases, multi-game) machines, fewer devices are needed to provide equivalent or better customer service, and most casinos are reducing their numbers of machines accordingly, including Foxwoods and Mohegan Sun, though not (yet) Twin River.

69 13 BIBLIOGRAPHY Christiansen/, Inc., et al., Legal Gambling in Connecticut: Assessment of Current Status and Options for the Future, January, Gerstein et al., Gambling Impact and Behavior Study: Report to the National Gambling Impact Study Commission, National Opinion Research Center at the University of Chicago, April 1, The Innovation Group, Video Lottery Terminals at New York Race Tracks: An Economic Study, June, Cummings, Will E., If You Build It, They Will Come -- But How Far? The Distance Factor in Regional Gaming Markets, a presentation to the 12 th International Conference on Gambling and Risk-Taking, Vancouver, B.C., May 30, 2003., Analysis of the Current Markets for Gaming in South Dakota, with Projections for the Likely Impacts of New or Enlarged Facilities, April 5, May be found at Analysis of Current Markets for Casino Gaming in Iowa, with Projections for the Revenues and Impacts of Potential New Facilities: Update, April 18, May be found at Slot Machines (or Video Lottery Terminals) at the Four Race Tracks of Massachusetts: An Opportunity for More Jobs and Tax Revenues, March 17, Cummings, Will E., Casinos Gravity According To Reilly Amended, a presentation to the 13 th International Conference on Gambling and Risk-Taking, Lake Tahoe, Nevada, May 25, May be found at Conference Web files/files/ Abstracts/index.htm., Assessment of the Value of a License for a New Casino in Davenport, Iowa, July 21, Cummings, Will E., Size Matters But How Much?, a presentation to the 14 th International Conference on Gambling and Risk-Taking, Lake Tahoe, Nevada, May 26, 2009., variety of reports to the Kansas Lottery Gaming Facility Review Board, 2008, 2009, and May be found at option=com_fabrik&view=table&tableid=4&calculations=0&resetfilters=0&itemid=63.

70 Market Analysis, with Projected Performance of a Slot-Machine Facility at Plainridge Racecourse Appendix: Details of the Gravity-Model Methodology Exhibits May 28, Jason Street, Arlington, Massachusetts Telephone: Fax: cummingsw@aol.com

71 List of Exhibits Exhibit A-1 Illustrative Distance Relationships (Mississippi) A-2 Distance Relationships II A-3 Distance Relationships III A-4 Portion of Model Inputs (2 pages) A-5 Gaming-Device Spending Ratios in Major US Markets A-6 Gaming-Device "Power Ratings" A-7 Slot Power Rating vs. Track Casino Retention % (Detail) A-8 Recent Projections Compared to Actual Results

72

73