When the Sun Sets on Net Metering: How the Cancellation of Net Metering Impacted the Potential Adoption of Residential Rooftop Solar Photovoltaics in Regina, Saskatchewan

Solar energy will play an important role in decarbonizing electricity generation and meeting global climate goals [1], [2], [3], [4]. While utility-scale solar offers zero-emissions electricity at a lower cost [5], rooftop solar photovoltaics (PV) will also contribute to achieving decarbonization goals. In their Net Zero by 2050 roadmap, the International Energy Agency (IEA) sets a milestone of 240 million households with rooftop solar PV installations by 2050 [2]. Jacobson et al. [6] out-line a 100% renewable energy scenario with over 1 billion residential rooftop solar PV units by 2050. The National Renewable Energy Laboratory (NREL) models rooftop solar PV scenarios ranging from 60 Gigawatts (GW) of capacity to 160 GW in the US by 2050 [7] and 8 to 23 GW in Canada by the same year [3], [8]. Policymakers hoping to achieve these targets will benefit from understanding how policy can encourage rooftop solar PV adoption.

Previous research has identified factors that influence rooftop solar PV adoption. In general, these factors fall into three categories: values, information, and financial returns.

With respect to values, households may adopt rooftop solar PV because they value self-sufficiency and want to reduce dependence on the electricity grid [9] or distrust their electricity utility [10]. Other households value inno-vation and technology and are eager to be early adopters of rooftop solar PV [11], [12]. Some may be motivated by environmental values, though this appears to be neither necessary nor sufficient for solar PV adoption [11]. Rather pro-environmental values may be related to a higher per-ception of personal benefits from solar [12]. The relative influence of values can vary at different levels of adoption. In a meta-analysis of literature on residential solar photo-voltaic adoption, Schulte et al. find that pro-environmental motivations are important for early adopters of solar, while financial incentives become more important as solar PV adoption spreads beyond the early adopters [13].

Information about solar energy can encourage solar PV adoption. This information can be the result of peer effects, where a household speaks with a friend or neighbour who installed solar [14], [15] or sees solar PV installations in their neighbourhood [12], [16], [17]. Confirming the importance of peer effects, Rode and Weber found that households in Germany were more likely to adopt rooftop solar if they saw solar installed in proximity to their homes [16]. Similarly, in a survey of 2065 Canadians, Parkins et al. find that re-spondents who regularly see solar panels are more likely to want to install panels on their own home [17]. Informa-tion can also be provided to households directly by a solar installation company [9] or a trusted third-party such as a non-profit organization or government agency [12]. In a study of solar adoption intentions in Ontario, Canada, Islam found that “Technology awareness and energy cost saving have a significant effect on the adoption probability, rein-forcing the need for effective education” about solar and its benefits ([18], p. 348). Learning about solar from a trusted source can provide reassurance that the technology works as intended [14] and can reduce informational barriers to solar PV adoption [19].

The most important factor influencing rooftop solar PV adoption appears to be financial return [19], [20]. Alipour et al. find that financial incentives that reduce solar purchase costs have a significant, positive impact on solar adoption [21]. Claudy et al. find that perceived cost savings from solar encourage adoption of rooftop solar PV [22]. Con-versely, several papers note that high upfront installation costs serve as a barrier to solar adoption [22], [23], [24]. Karakaya & Sriwannawit review barriers to solar PV adoption and find that households evaluate the financial return provided by solar relative to alternatives; access to low-cost, non-solar energy alternatives discourage solar adoption [9]. In gen-eral, Kastner & Stern find that the financial consequences of an investment such as installing residential solar is rela-tively more important than factors such as the “disposition” or environmental values of a household [25]. They also find that government funding for investments increases the likelihood of investment, and grants may be more effective than low-interest loans [25]. Williams et al. [20] create a parsimonious model of solar PV adoption and find that the net present value (NPV) of the solar installation to the owner explains variations in solar PV adoption across countries. We adopt the parsimonious Williams et al. [20] approach in this study and focus on NPV and the financial returns generated by rooftop solar.

Government policy can impact NPV by influencing the capital cost of solar PV installations, the financing cost of the project, the value of avoided electricity purchases during times of self-consumption, and the value of excess power sold to the grid in times of surplus. Physical variables like the quality of the solar resource also influence NPV [26]. Multi-story buildings in high-density neighbourhoods may have limited roof space for solar relative to electricity de-mand [26]. Shading lowers the potential output of solar and reduces NPV. We consider government policy, the solar resource, roof space, and shading in this study.

In this paper we contribute to the literature in three ways. First, we assess how the net present value of rooftop so-lar PV is impacted by government policy decisions using real-world data on solar irradiation, rooftop space and ori-entation, and electricity usage patterns. Second, we use survey data to understand the level of financial return that households require to adopt rooftop solar PV. Third, we combine our analysis of the financial returns made possi-ble by rooftop solar with our survey data to estimate upper bounds for solar adoption. We can then assess the de-gree to which policy can influence the installation of solar in our case study community of Regina, Saskatchewan. Our mixed methods approach is novel in the solar adoption lit-erature. While papers such as Dargouth et al. [27] have analyzed the financial returns to solar under different policy scenarios, they did not link these to survey data of potential adopters. The bulk of the solar adoption literature focuses on either post-hoc solar installation data or surveys asking whether households intend to install solar. We offer a real-world analysis of the implications of moving away from a net metering policy.

We focus on Regina because there is an opportunity for solar to reduce GHG emissions in the city, and because pol-icymakers in Regina are seeking to encourage renewable energy supply. There have also been recent changes to so-lar policy design in Saskatchewan that impact the financial returns of solar PV.

Regina is the capital city of the province of Saskatchewan and boasts some of the highest solar irradi-ation values in Canada [28]. Regina’s electricity is supplied by provincially owned Crown utility SaskPower, which re-lies on natural gas (42%) and coal (37%) power generation for most of its electricity supply [29]. The greenhouse gas (GHG) emissions intensity of SaskPower’s electricity was 637 tonnes carbon dioxide equivalent (CO2e) per Gigawatt-hour (GWh) in 2021 [29]. Installing solar in Regina offsets this relatively high-emissions electricity.

The municipal government in Regina has committed to achieving a 100% renewable energy future by 2050 [30]. Our analysis seeks to inform the City of Regina’s efforts as they create a strategy for achieving this goal. Programs to encourage rooftop solar PV adoption can be part of an integrated approach to municipal emissions reductions, that also involves electrification of vehicles and buildings [31].

There have been a series of changes to solar policy in Saskatchewan in recent years. SaskPower formerly offered a per-kilowatt (kW) capital subsidy for solar PV installa-tions and allowed customers to enroll in a net metering program. Like other net metering programs, SaskPower’s program compensated solar electricity exported to the grid at the retail electricity rate. After conducting a public en-gagement process in 2017 [32], [33], SaskPower refined its residential net metering program to offer longer term con-tracts to customers who installed solar PV. The program changes encouraged more solar adoption in the province and led to an increase in business for the Saskatchewan solar industry. In September of 2019, SaskPower stopped accepting applicants into the net metering program, stat-ing that they had met the 16-Megawatt (MW) quota for the program. SaskPower eliminated the per-kw subsidy and switched new solar customers to a ‘net billing’ program that compensated electricity sold back to the grid at 7.5 cents (Canadian dollars, CAD) per kilowatt-hour (kWh) rather than the residential retail rate, which was 14.2 cents (CAD)/kWh at the time. This was met with criticism from the solar in-dustry and led to a 90% decline in solar installations in the province within a year of the program change [34], [35]. In May 2021 the Canadian federal government introduced a program to encourage energy efficiency and rooftop solar PV. The Greener Homes Grant provides rebates of $1000 per kilowatt (kW) of installed solar, up to a maximum of 5 kilowatts per household [36], [37]. In this paper we assess the impacts of this policy change and offer recommendations on program design to increase solar adoption.

We assess the impact of the changes to Saskatchewan so-lar policy by simulating the solar output and financial returns that can be achieved in Regina under three policy scenar-ios: net metering, net billing, and net billing with the Greener Homes Grant. We use GIS analysis to identify suitable roofs in Regina and assess any shading that may occur. We use solar irradiation and temperature data to calculate hourly ca-pacity factors for simulated solar installations in Regina [28]. We obtain anonymized smart meter data from Saskatoon, SK to understand the potential shape of electricity demand in Regina homes [38], [39]. We match Saskatoon and Regina neighbourhoods to ensure load shapes reflect differences in the age of housing and demographic characteristics. We then combine roof and load data to simulate over 4 million potential roof-load combinations and calculate NPV and net monthly return for each combination.

We are interested in how changes to the financial return of rooftop solar impact solar adoption in Saskatchewan. To assess willingness to install solar at different levels of financial return we conduct a telephone survey of 451 Regina residents. We ask solar-eligible households what financial return they would require in order to install rooftop solar PV in their homes. We combine these responses with our solar generation simu-lations to conduct a bootstrap analysis that identifies the upper limit for solar adoption under our three policy scenarios.

We assess residential rooftops in Regina, Saskatchewan for the physical potential to install solar. Hourly solar irradia-tion and temperature data for 2017 is sourced from Environ-ment Canada’s Canadian Weather Energy and Engineering (CWEEDS) dataset [28]. We use the approach outlined in Masters [40] to calculate hourly solar capacity factors. We pay particular attention to the heterogeneity of rooftops when it comes to south-facing roof area, angle of the roof relative to due south (i.e. collector azimuth), and shading.

Hourly rooftop shading factors for Regina are found using open data geographic information system (GIS) files and a free, open-source GIS software (QGIS). We use these data sources and software for ease of access and replicability of results. Required GIS input files are a digital surface model (DSM) in raster format [41] and building footprints in vector format [42]. An additional data file containing hourly solar azimuth and elevation is also needed [43].

To find hourly shading factors, we filter the rooftops by slope and aspect (orientation). Both slope and aspect rasters are created from the DSM raster using built-in QGIS functions of ‘Slope’ and ‘Aspect’. From this, suitable rooftop areas are filtered based on the following criteria:

• Raster files overlapping with building vectors;

• South facing and/or flat (orientation [44] between 90◦ and -90◦ and/or slope less than 5◦); and

• Not a vertical or near vertical wall (slope less than 70◦). We create shading rasters for each hour of daylight over the entire city using the DSM raster, the solar azimuth and elevation, and the ‘Hillshade’ function, with the z-factor (ex-aggeration factor) set to 1 to mimic real shading conditions. These hourly rasters are then clipped to only the suitable rooftop areas found. Using the ‘Zonal statistics’ function, the average hourly shading function is found for each suitable rooftop area, as well as the size of the horizontal projection (footprint) of the suitable rooftop area. Finally, the actual size of the suitable rooftop area is calculated based on the size of the horizontal projection of the suitable rooftop area and the associated slope.

Within this process, we make two key assumptions due to data availability and computational processing capabilities, re-spectively. First, within the DSM raster, most vegetation (tree) coverage that overlapped building footprints is assumed to be filtered out based on the suitable rooftop criteria, as surface coverage from vegetation would not fit the criteria of being flat, nor primarily south facing. This is confirmed through visual inspection of the resulting raster files. Secondly, we calcu-lated the shading factors for only the solstices and equinoxes daylight hours. Based on the historical trends in azimuth and elevation changes, as well as sunset and sunrise times, the rest of the year is linearly interpolated from the solstice and equinox shading data.

Our database contains 101,128 residential, commercial and industrial buildings in Regina. We focus on residential rooftops and so select only those properties in areas zoned for medium and light density residential. Zoning data is collected from the City of Regina [42]. We also filter out 1832 multi-unit proper-ties and focus on single-detached and duplexes to match our survey design. This leaves us with 88,876 unique residential buildings. After filtering out buildings with missing information we are left with 78,630 residential buildings in Regina, mean-ing we have complete information for 88.5% of the unique residential buildings we identify. For each of these buildings we know the south-facing area of the roof, the angle of the south-facing roof relative to due south and have hourly shad-ing factors for each daylight hour of the year (4501 hours in all) using the approach outlined above.

We calculate hourly capacity factors (CFit) for one year (8760 hours) for each of the 78,630 rooftops based on 2017 solar irradiation and temperature data (see Equation 1) [28]. Tem-perature data is relevant because temperature impacts the efficiency of solar photovoltaic production; panels produce electricity more efficiently at lower temperatures. We correct for the temperature of the solar cells using the approach out-lined in Masters [40]. We assume a nominal operating cell temperature (NOCT) of 46◦C based on a review of available solar panels.

We account for the angle of the roof relative to due south in our calculations (the collector azimuth angle φC in Masters, 2004). We then apply shading factors that pro-portionately reduce the capacity factors in each hour. The shading factors (S_it) represent the proportion of each roof (i) that is shaded in any given hour (t). We assume that solar panels will take up a maximum of 80% of the roof area, leaving room for maintenance access. We also assume that solar panels will be placed strategically on each roof to mitigate shading and reduce our shading factors by 30%(0.3) to account for this shade mitigation [45].

The shading correction is then applied using Equation 1:

where, CFitu is the unshaded capacity factor, and CF_it is the capacity factor that results after accounting for shad-ing [46].

Some of the policies to encourage solar PV adoption differ-entiate between electricity that is exported to the grid and electricity that is consumed ‘behind the meter’ within the property on which the solar is installed. To account for these differences, we calculate measures of “self-consumption”. This is the proportion of electricity consumed by the res-idential homeowner. Hourly load data is not available for Regina, Saskatchewan because smart meters are not yet widely deployed. Instead, analog electricity meters are read occasionally and consumption between meter read-ings is estimated. The situation is different in Saskatoon, Saskatchewan where the local utility Saskatoon Light & Power has deployed smart meters throughout their service area. We use hourly load data from Saskatoon smart me-ters as a substitute for the unavailable Regina load data [38], [47]. We believe the substitution is valid since Saskatoon and Regina are located 250 kilometers apart, face very sim-ilar climates, and are home to people with similar lifestyles. To ensure the load shapes from Saskatoon closely approxi-mate expected load shapes in Regina we match neighbour-hoods in each city based on socio-demographics and hous-ing characteristics. We then calculate self-consumption factors for each roof and load pair within each matched neighbourhood set.

We compare the distribution of capacity factors using the ‘ggridges’ package in R [48]. This visualization approach indicates the relative frequency of capacity factors within our simulations. Capacity factors are listed along the x-axis and the height of the curve indicates the relative frequency of each capacity factor score. The highest points on the dis-tribution curves indicate the most frequent capacity factor scores. Similar visualizations are used for self-consumption scores, net present value (NPV), and monthly net returns (see figs. 1 to 9 below).

The shape of electricity load will vary by household and is determined by the lifestyles of the occupants, the nature of the appliances in the home, and building characteris-tics such as energy efficiency, heating system, and water heating energy source. To match Regina and Saskatoon neighbourhoods we compare six neighbourhood statistics from the 2016 Census [49], [50], [51]:

• Median after-tax income of households in 2015 (Y_l);

• Proportion of population with postsecondary certificate, diploma or degree (E_l);

• Proportion of owner-occupied dwellings (O_l);

• Proportion of dwellings built pre-1960 (A_l);

• Proportion of single-detached homes (S_l); and

• Average monthly shelter costs for owned dwellings ($)(C_l).

We pool the neighbourhood statistics from Regina and Saskatoon and normalize all the variables using Equation 2:

where, l is the neighbourhood, and k represents each of the six statistics mentioned above. We then calculate the Eu-clidean distance between each of the Regina and Saskatoon neighbourhood pairs using Equation 3:

where, l represents Regina neighbourhoods and m rep-resents Saskatoon neighbourhoods. We then sequentially select Regina-Saskatoon neighbourhood pairs with the low-est Euclidean distance scores. In all we match twenty-eight Regina neighbourhoods with Saskatoon neighbourhoods that possess similar characteristics.

We use solar capacity factor data for the 78,630 Regina rooftops, and 1648 unique load profiles from Saskatoon homes to calculate solar self-consumption. We calculate self-consumption for each roof-load combination by joining the annual capacity factor data for each Regina roof with each of the load profiles in the matched Saskatoon neigh-bourhood file. This means for a given Regina neighbour-hood with i roofs, and v matched Saskatoon load profiles we have i ∗ v solar-load combinations.

We scale the size of the simulated solar installation to equal the size that would produce annual electricity equal to annual household load, or the maximum size possible on the available roof area, whichever is greater. We assume that 6m² of roof area is required for each kilowatt of solar capacity.

We then compare hourly solar output in the simulated solar installation to hourly household load. When solar output is greater than load in a given hour, we calculate the amount of electricity exported to the grid (e.g. see Fig- ure 1 where solar output is greater than load during peak solar production hours). We then calculate annual solar out- put (output_iv), load (load_v), and electricity exported to the grid (export_iv). Self-consumption (SC_iv) for each roof-load combination is calculated using Equation 4:

Where i refers to Regina roofs, and v refers to Saska-toon Light & Power load profiles. In total we calculate self- consumptionstatisticsfor4,481,346uniqueroof-loadpairs. We summarize the distribution of self-consumptionpropor- tions using the ‘ggridges package in R [48].

We simulate three solar policies modelled on past and present policies in Saskatchewan:

• Net metering policy that allows customers to receive credit for excess electricity sold to the grid at the 2022 retail rate of 14.705 cents/kWh (plus sales tax);

• Net billing policy that provides customers with a credit for excess electricity sold to the grid at a discounted rate of 7.5 cents/kWh;

• Net billing combined with a capital grant modelled after the existing Greener Homes Grant offered by the Canadian federal government that offers $1000 per kilowatt of installed solar, up to a maximum of $5000 [36].

We model the outcomes of each policy with respect to the net present value (NPV) of the solar installation, and themonthlyfinancialoutlookforahouseholdinstallingsolar. Williams et al. [20] suggest that NPV provides a parsimo- nious model of solar adoption. They find empirical support relating NPV to solar adoption rate. We calculate NPV us- ing the formula outlined Williams et al. [20] and presented in Equation 5:

Where:

C_total: capital cost of the PV system ($)

S: subsidy for PV purchase ($)

output_iv: total electricity produced in one year by the solar installation (kWh)

SC_iv: self-consumption share (%)

RP: retail price of electricity ($/kWh)

Solar P rice: price paid for solar energy exported to the grid ($/kWh)

µ inflation rate (%)

r: social discount rate (%)

T : terminal year of the solar installation which is 25 assum-ing a 25-year life for the solar panels.

The cost of residential solar PV in Regina, Saskatchewan was \$2300/kW installed in 2019 [52]. This per-kW cost is multiplied by the size of the solar installation for each roof-load combination to calculate installed capital cost. Residential solar installations are also charged flat fees of \$315 for an interconnection study, \$498.75 for a bidi-rectional meter, and must pay \$150 to obtain an electrical permit. These start-up costs are added to the capital costs for each installation we simulate.

The retail price of electricity for residential customers in Saskatchewan was 14.705 cents/kWh in 2022 for energy plus a carbon pricing charge of 0.6393 cents/kWh [53]. On top of these volumetric charges, a sales tax of 5% is levied, along with a municipal surcharge of 10%. In total the retail electric charge in 2022 was 17.0974 cents/kWh. The price paid for solar energy exported to the grid is 7.5 cents/kWh in the current residential solar program [54].

We set the inflation rate equal to 0% and deal in real values, assuming that electricity prices will rise at the rate of economy-wide inflation. We use 3% for our social dis-count rate in our main results, and then explore sensitivity to higher discount rates in other simulations. For each policy we provide a graphical summary of the distribution of net present values (NPV) using the ‘ggridges’ package in R [48].

We also calculate the monthly net cost of installing so-lar in each roof-load combination. This monthly format is created to simulate the financial picture that a household would face when deciding whether to take out a loan to finance a solar installation. Calculating the monthly net cost also allows us to compare our findings against the results of our telephone survey of Regina residents. To calculate monthly cost we amortize the cost of the solar installation over the 25-year life expectancy of the panels. We calculate a monthly interest rate (im) using Equation 6:

Where ia is the annual interest rate, which we set at 3%to represent the cost of borrowing. We amortize the cost of the solar installation, including the start-up cost, using the standard amortization formula outlined in Equation 7:

Where n represents the term of the loan in months, which we set at 300 (25 years × 12 months in a year).

We designed a survey instrument to ask Regina residents about their household eligibility and willingness to install solar (see Appendix A). The survey was carried out between Au-gust and September 2019. Regina residents were contacted via landline and cellular phone numbers. We surveyed 451 Regina residents, and ensured that thresholds were met for demographic representation. In particular, we did not want to weigh any one demographic group by more than two times. For example, if Regina residents between twenty and thirty years of age represented 10% of the population, we wanted to ensure they represented at least 5% of our sample.

Participants were asked a series of questions related to their support for City of Regina policies such as the goal of becoming 100% renewable by 2050 [55]Participants were then asked specifically about solar energy. To establish their eligibility to install solar we asked participants what type of home they lived in and whether they owned their home. Those who answered that they live in a single detached house or an attached house (townhouse, duplex, triplex), and who own their homes were asked follow-up questions. Our reasoning is that apartment dwellers are not individually able to install solar panels, and instead need to work through a structure like a condominium board. Renters likewise do not have the agency to install solar panels on their residence. These are the same eligibility criteria used to model distributed generation in the North American Renewable Integration Study (NARIS) [3].

Follow-up questions focused on suitability of roofs for in-stalling solar panels and willingness to install solar panels (see questions 20-25 in the survey instrument in Appendix A). Relevant for this analysis is our use of a payment ladder to assess how willingness to install solar was influenced by the monthly net return or net cost (Q24). We first described the following scenario to participants:

“Imagine you borrow money from a bank to purchase and install solar panels. You make monthly payments to repay the loan, but installing solar panels also reduces your electricity bill. What best describes your desire to install solar panels:

• I would install solar panels even if my total monthly costs increase;

• I would install solar panels if I broke even;

• I would install solar panels only if my total monthly costs decreased.”

Participants who answered that they would install solar if they broke even were assigned an adoption threshold of$0/month in total cost change. For those who responded that they would be willing to adopt solar even if total monthly costs increased, or only if total monthly costs decreased, we asked a payment ladder question (see Figure 2). Participants were assigned a random starting value on the payment ladder and asked whether they would install solar panels if they received that value in monthly net financial costs or savings. An answer of ‘Yes’, indicating a willingness to install solar at the stated amount, led to a follow-up question ‘up’ the payment ladder. An answer of ‘No’, indicating that they would not install solar at the stated amount, led to a follow-up question ‘down’ the payment ladder. These questions continued until the answer switched from Yes to No, or No to Yes. A solar adoption thresh-old was then assigned based on where the participant landed on the payment ladder.

We use these responses to construct a demand curve representing willingness to adopt solar at various levels of net monthly return.

We estimate the adoption of solar panels that would result from each solar policy by comparing the distribution of net monthly returns that could result from each policy with the required financial return thresholds gathered from the survey of Regina residents. For each policy we create a bootstrapped sample by selecting a random observation from the net re-turn vector (the roof-load simulations), and matching it with a random sample from the survey results. Since we do not know the conditional probability distribution that would tell us how the required return on solar (solar threshold) varies by the solar potential of a given home, we assume an equal probability that a given simulated roof-load combination could be home to a given survey respondent. We repeat this ran-domized matching procedure 100,000 times for each policy, and then calculate the number of households where the finan-cial return that could be generated by the simulated roof-load combination exceeds the required solar return threshold. We then assume that these households would adopt solar given complete information about the solar potential on their roof, the financial implications of a given solar policy, and their own required financial return from solar. This is an upper bound for solar adoption since complete information would be unlikely for most households.

Figure 3 shows an overhead view of a sample of the Regina rooftops included in our analysis. South-facing roofs appear in bright yellow and register the greatest potential for solar electricity production. Smaller shapes on a given lot repre-sent garages or sheds. Accounting for the heterogeneity of rooftops in Regina allows us to estimate a distribution of solar PV capacity factors. In clear sky conditions, without any shading, a solar PV installation could be expected to perform with an annual capacity factor (CF) of around 18%in Regina, SK (“CF unshaded” in Figure 4). Variations in these unshaded capacity factors results from variations in the collector azimuth angle for each simulated rooftop.

Shading on a rooftop reduces photovoltaic production. Shading could be created by trees or neighbouring build-ings. Once we account for shading, the performance of rooftop solar in Regina is reduced. Accounting for shading, the most frequent annual capacity factor simulated in our 78,530 rooftops is just under 15% (“CF shaded” in Figure 4).

The financial return generated by rooftop solar depends on policy design, geographic factors such as solar irradi-ation, roof orientation, and shading, and technical factors such as photovoltaic efficiency. The temporal relationship between solar output and household electricity load is also important for policies like net billing. Under a net billing program, when solar is consumed within the home it offsets the residential price of electricity. When solar is exported to the grid, a lower value credit is offered to the solar producer. Households that match their electricity use to solar energy production earn greater returns under a net billing program by reducing their purchases of electricity from the grid.

In our simulations, we find a range of self-consumption factors centered on a median value of 44% and a mean of 53% (see panel A in Figure 5). We treat the shape of home electricity load as exogenous, meaning that it doesn’t shift in response to policy. In practice, net billing would provide an incentive for homeowners to shift electricity use to the sunniest parts of the day when they are producing the most solar energy. This will become increasingly important as the vehicle fleet is electrified; vehicle charging may occur dur-ing times of peak PV output to maximize self-consumption. Our assumption that households don’t shift electricity usage means the self-consumption numbers are a lower bound on what would be achievable.

Figure 5 Panel A also displays a high frequency of roof-load pairs with self-consumption proportions closer to 100%. This typically occurs when roof size constrains the size of solar installations. Panel B in Figure 5 presents a scatterplot of self-consumption proportions compared to roof size. We add a smoothed conditional mean to show the relationship between roof size and self-consumption. This line trends downwards showing higher self-consumption on smaller roofs. In cases where self-consumption shares are close to 100%, roof size is around 10m2. This would represent the roof of a small garage or large shed. In these cases, solar panels generate only a small share of household elec-tricity consumption, and nearly all of this solar electricity is consumed at the time it is generated. This would mean that these small installations largely offset the purchase of electricity from the grid. So, while smaller roofs lead to less solar production in total, they can achieve higher value per kilowatt-hour under a net billing program by offsetting high-cost grid electricity.

Using the parsimonious equation introduced in Williams et al. [20], we calculate net present value (NPV) for solar in-stallations under three policy scenarios: net metering; net billing; and net billing with the Canadian federal government’s Greener Homes Grant [36]. Figure 6 presents the distribution of NPV values under each policy. Table 1 summarizes the NPV results.

SaskPower’s former net metering program offered the most generous payment for solar energy in Saskatchewan. Under that program, over 95% of our simulated roof-load com-binations would have earned a positive NPV over the 25-year life of the solar installation. The mean value of the NPV under net metering is \$5062 and the median is \$4440. Calculated us-ing the mean NPV, this program offered an equivalent annual net benefit of \$292 on average [57], [58].

When the net billing program was introduced in Septem-ber of 2019, the potential NPV of solar projects dropped. Net billing meant that 70% of projects would achieve a positive NPV over the 25-year life of a solar installation. Financial returns fell to a mean NPV of \$984 and median of \$659. The equivalent annual net benefit was reduced to \$56/year. This drop in financial return explains why solar installations fell by 90% after the program change [35].

The Canada Greener Homes Grant helps to restore the financial return of residential solar in Regina. The combina-tion of net billing and the Greener Homes Grant can provide a positive NPV for 92% of our simulated roof-load combina-tions. The mean NPV is \$3,990, and the median NPV is \$3,837, which approaches the value of the former net me-tering program. Equivalent annual net benefit calculated for the mean NPV is $229 per year, which is about \$19/month. The Greener Homes Grant program does introduce some additional logistical barriers for homeowners. To receive the incentive homeowners must first participate in an energy ef-ficiency audit. Eligible homes can then receive an incentive of \$1000 per kilowatt of installed solar, up to a maximum of \$5000 in cash incentives. The Greener Homes Grant can also only be applied to solar PV systems equal to or greater than 1 kW [36], which explains why a relatively larger subset of solar installations remain in the negative NPV territory in Figure 6.

The NPV analysis outlined above assumes a discount rate of 3%. This relatively low discount rate could be understood as either a social discount rate, or the cost of capital for those who install solar. In practice, many households have higher discount rates when it comes to evaluating the desirability of rooftop solar. Talevi [59] and De Groote and Verboven [60] estimate that discount rates may be as high as 15% when evaluating rooftop solar opportunities. This indicates that households heavily discount the stream of future electricity savings that will arise from installing solar panels. Using a discount rate of 10%, Figure 7 summarizes the NPV of solar under each program. Table 2 presents summary statistics at a 10% discount rate (Under a discount rate of 15%, almost no roof-load combinations would achieve a positive NPV).

The net metering program offers a positive NPV for only 1% of our roof-load combinations at a discount rate of 10%. This is close to the proportion of households that had installed solar panels at the time of our survey (see Section 3.5 be-low). No homes in Regina would earn a positive NPV with the net billing program at this high discount rate, which helps to explain why residential rooftop solar installations have fallen sharply since SaskPower switched from net metering to net billing. Interestingly, the Greener Homes Grant would lead to a positive NPV for 21% of our simulated homes. This is because the grant is received upon installation of the panels. This upfront value is not discounted and so has a greater impact on NPV than a stream of future revenue receipts.

Homeowners may not think about solar investments in terms of the lifetime NPV of the installation. On a more immediate timescale, homeowners will want to know how the benefits of solar compare to the cost of financing the solar installation. We calculate the monthly cost of financing a solar installation under each of the three policy scenar-ios. We then subtract this amortized monthly cost from the monthly electricity bill savings achieved by installing solar to calculate monthly net financial return. Figure 8 presents the distribution of monthly net returns. The shape mirrors that of Figure 6 above. Mean monthly net returns are $25 for net metering, \$5 for net billing, and \$19 for net billing with the Greener Homes Grant (see Table 3). Median net returns are \$22 for net metering, \$4 for net billing, and \$19 for net billing with the Greener Homes Grant (Table 3).

In Section 3.6 we compare these monthly net returns to the net return thresholds identified in our survey of Regina residents to estimate solar adoption in Regina.

Of our 451 survey respondents, 75.9% live in a single-detached or attached house that they own. We treat this group as the population who have the agency to install so-lar on their own properties. A smaller percentage of survey respondents (53.5%) meet this agency criteria and have a south-facing roof where solar could be installed. We treat those who own their own single-detached or attached house, and have a suitable south-facing roof, as the population that is eligible to install solar on their homes (the solar-eligible population).

Only eight households in our survey, or 1.8%, already have solar installed on their homes. Of the survey respondents, 13% stated that they would likely or very likely install solar in the next 24 months. Note that our survey was carried out in August and September of 2019, just before the change to SaskPower’s net metering program was announced.

We asked survey respondents who met our eligibility crite-ria follow-up questions about the financial return that would incentivize them to install solar. As a proportion of solar-eligible households, 40.6% stated that they required their total monthly cost to decrease; i.e. they required a positive monthly net return (see Table 4). Another 33.1% of solar-eligible house-holds would be willing to install panels if they broke even; i.e. net monthly costs would be equal to net monthly electricity bill savings. A smaller proportion of solar eligible households (12.3%) were willing to install solar even if their monthly costs increased. Still others stated that they would not install solar in any circumstances (11.7% of solar-eligible households) and a small portion (2.3%) declined to answer the question. If a participant stated that they would install solar only if their total monthly costs decreased, we followed up by asking about willingness to install solar if a specific financial return was achieved. We did the same for those who said they would install solar even if their costs increased. These par-ticipants moved up and down the payment ladder illustrated in Figure 2 until settling upon a net monthly return (positive or negative) that would lead them to install solar. Partici-pants who would be willing to install solar if they broke even were coded as requiring a net monthly return of $0 in order to install solar.

Figure 9 displays the proportion of Regina households that would install solar at various levels of net return. The x-axis is cumulative since we assume, for example, that if a household would install solar when a net return of $20 was achieved, they would also be willing to install solar if a net return of $40 is achieved. The biggest jump occurs between a monthly net return of -$5 per month and $0 per month. Those who answered that they would be willing to install solar if they broke even were added to the curve at that point.

We use our survey responses, and the net monthly re-turn data from our rooftop analysis to estimate potential solar adoption by solar policy. Because we do not have a dataset where rooftop solar potential is matched to survey responses, we create 100,000 randomly matched roof-top and survey responses for each policy. Our bootstrap anal-ysis results are summarized in Table 5. The second col-umn indicates the proportion of our bootstrapped sample for which net return according to the characteristics of the matched rooftop would exceed the required financial return from the matched survey response. With higher monthly net returns, the net metering program achieves the highest proportion of solar-eligible households with net monthly re-turns greater than required thresholds at 65%. Net billing achieves 48% of matches where net monthly return exceeds the threshold, while net billing plus the Greener Homes Grant achieves 61%. These proportions represent the pro-portion of solar-eligible households who would consider installing solar, and solar-eligible respondents made up 46% of the participants surveyed. When we scale by 46%, we see potential solar adoption levels of 30% for net me-tering, 22% for net billing, and 28% for net billing plus the Greener Homes Grant. These lower percentages repre-sent the proportion of all Regina households who would potentially adopt solar.

Interpreting these proportions requires a note of caution. We can think of the percentage of ‘potential adopters’ as the upper limit for households that would install solar. We assume that the potential adopters would still require en-abling conditions before they install solar. These conditions include:
• complete information about the cost of solar and the value of electricity savings benefits [19]

• confidence electricity savings benefits would be re-ceived over the lifetime of the solar installation;

• access to financing at a 3% interest rate.

If these conditions are missing, households may not install solar. Access to financing could be a problem partic-ularly for low-income households.

Based on the upper limit proportions in the ‘Potential adopters’ column, we calculate the amount of solar capacity that could be installed on residential homes in Regina, and the amount of electricity that would be generated each year. Average solar size and average solar capacity factor are based on the characteristics of the bootstrapped sample. We take the mean of installation size (kW) and capacity fac-tor (%) for those households where net return exceeds the required threshold. The amount of solar energy that could be generated from rooftop solar ranges from 99 GWh/yr to 129 GWh/yr.

As the City of Regina works towards a goal of using 100%renewable energy by 2050, residential rooftop solar can play a role. Regina, Saskatchewan boasts a strong solar re-source, and our analysis finds that residential rooftops can generate electricity with a capacity factor of around 15%given existing technology and after accounting for shading. On the high end of potential adoption, residential rooftops in Regina could be home to nearly 100 MW of rooftop solar capacity and generate 129 GWh of electricity per year. The City of Regina estimates that city operations use 230 GWh of energy per year in all of its forms (i.e. electricity, heating, vehicles) [62].

To generate half of that energy with residential rooftop solar electricity, Regina would need approximately one-third of Regina households to install solar photovoltaics on their roofs.

To meet a city-wide goal of 100% renewable energy by 2050, Regina will need to develop a broader mix of electricity generation options. Looking only at electricity, Regina homes, businesses and industry use approximately 2500 GWh of electricity per year. Much more electricity will be required as Regina shifts to electric vehicles, electric heat, and electric industry to reduce greenhouse gas emis-sions. This means that while residential rooftop solar can contribute to achieving Regina’s 100% renewable goals, other energy sources will be necessary to meet energy demand. Additional solar photovoltaic development could include more rooftop solar on commercial buildings within Regina, as well as large, utility-scale (i.e. >10 MW) solar farms outside of the city. South-central Saskatchewan is also an ideal location for wind energy developments [63].

The City of Regina could look to regional partnerships to boost production of wind and solar energy. For example, the Cowessess First Nation has developed a wind-solar-battery energy production site to the east of Regina [64]. Further partnerships with First Nations and rural municipalities can allow Regina to secure the renewable energy needed to achieve the 100% renewable goal.

Policy design impacts the desirability of installing solar in Regina. Of the policies we evaluate, net metering offers the highest net present value and net monthly financial return to homeowners in Regina, SK. SaskPower had a net metering program in place until 2019 and it was popular. That popu-larity led to the program’s cancellation once SaskPower hit a cap of 16 Megawatts (MW) of solar installed under the net metering program [34]. Our results confirm that the net billing program that replaced net metering offers lower finan-cial returns. The move to net billing decreased the mean NPV of rooftop solar from $5062 to $984, while the mean monthly net financial return fell from \$25 to \$5 at a discount rate of 3%. This drop makes rooftop solar only marginally de-sirable for most households. Any unplanned maintenance or excess shading could put a net financial return into jeopardy. This drop in the financial benefit of rooftop solar helps to explain why solar installations in Saskatchewan decreased by 90% after the program change [35].

The federal government’s Greener Homes Grant, when combined with SaskPower’s net billing program, restores the NPV and monthly net financial return of solar to levels nearly equal with net metering. The federal government will provide households with an incentive payment of \$1000 per kilowatt of solar photovoltaic capacity installed, up to a maximum of \$5000 [36]. The combination of net billing with the Greener Homes Grant increases the mean NPV of solar to \$3,990 and the monthly net financial gain to \$19 at a discount rate of 3%. If the City of Regina would like to increase adoption of rooftop solar it could consider topping up this capital incentive with an additional payment from the municipality.

Residential rooftop solar in Regina may not approach the upper limit of its potential if homeowners have high dis-count rates and policies do not account for this. Without declines to the cost of installing solar, households with dis-count rates of 10% or 15% will likely not install solar, even with Regina’s relatively good solar irradiation conditions. We find that the mean NPV for solar is negative for solar under all three policy designs when discount rates are set to 10%. When discount rates are high, the upfront capital grant provided by the federal Greener Homes Grant pro-gram does improve the NPV of solar relative to programs that do not have a capital grant. With a combination of net billing and the Greener Homes Grant, 21% of households would have a positive NPV even when discount rates are set at 10%. This is an improvement over the net meter-ing program which provided a positive NPV to only 1% of households at a discount rate of 10%. This finding speaks to the advantage of grant-based programs in countering the myopia of homeowners. As Talevi [59] notes, policymakers who wish to encourage solar installations can design in-centive structures that frontload payments for solar. These upfront payments can encourage solar at a lower cost than programs that focus on paying higher electricity rates in the future.

Low-interest loan programs may also be of value to re-move barriers to rooftop solar and reframe benefits in terms of monthly net financial returns. As reported in several other papers, lack of financing and high upfront costs can be a barrier to the installation of rooftop solar [22], [23], [24], [25]. A capital grant plus access to a lower-interest loan can enable invest-ments without high, upfront expenses. Just as households discount the stream of future benefits of an investment in rooftop solar, they may discount the stream of future costs of loan repayments. This would turn attention to the monthly net financial return from solar, which is generally positive under all three policy scenarios (see Table 3). The City of Regina could explore provision of ‘Property Assessed Clean Energy’ (PACE) loans for rooftop solar to provide low-interest loans. These loans are attached to properties, rather than individuals, which can also alleviate household concerns that they may move before recouping the value of their investment in rooftop solar. In general, solar adoption will be higher when efforts are made to spread costs over a longer period of time, and shift benefits to the present.

With high irradiation values, and cold, clear winter days that allow for high photovoltaic efficiency, Regina, Saskatchewan is ideally situated for solar energy. We find that residential rooftop solar can achieve capacity factors of 15% on aver-age in Regina once shading is considered. Assuming that the size and shape of residential electricity load is compa-rable to Saskatoon, the average household in Regina uses close to 7000 kilowatt-hours (kWh) of electricity per year. A solar installation of about 5.3 kilowatts (kW) would generate electricity equivalent to this electricity load.

We find that the switch from net metering to net billing led to a sharp decline in the NPV of rooftop solar in Regina. This policy change reduced the potential rooftop solar adop-tion rate from 30% of Regina households to 22%. We find that the Greener Homes Grant restored the NPV of rooftop solar in Regina to levels nearly equal to the net metering program. The Greener Homes Grant also has the advantage of providing an upfront incentive, which may be more effective at encouraging solar PV adoption due to the high discount rates of households [59], [60]. The City of Regina can look to build on the Greener Homes Grant as a method of further encouraging rooftop solar PV adoption in Regina. The City of Regina can also consider providing low-interest loans for rooftop solar installations to remove the barrier of high upfront capital costs.

In our analysis, we focus on the perspective of the home-owner considering the adoption of solar. We do this to estimate upper limits on what solar could contribute to the City of Regina’s 100% renewable energy goals. We do not forecast solar adoption into the future. Follow-up work could calibrate solar adoption using historic solar installation data and use that to produce a forecast of solar adoption [65], [66]. This modelling could also incorporate the peer impacts of solar adoption; households are more likely to adopt solar if they see that their neighbours have done so [12], [14], [15], [16], [17], [65]. Future analysis could also consider the broader social value of residential rooftop solar. Electric utilities through-out North America are grappling with the design of rooftop solar compensation policies. From the perspective of the electric utility SaskPower, solar energy is worth only the avoided cost of reducing natural gas usage. When the sun is shining, SaskPower can run its natural gas power plants at a lower capacity and save on fuel costs. Viewed this way, SaskPower values solar electricity at only $0.04 per kWh, plus any associated carbon pricing savings [32]. Even the net billing program could be perceived as overly generous from the utility’s perspective [67]. SaskPower has shown a preference for large, “utility-scale” solar installations that can achieve generation costs closer to SaskPower’s avoided cost [68]. The utility also plans to build much more wind capacity in coming years since the levelized cost of wind is now equal or less than SaskPower’s avoided cost of gener-ation [5], [69], [70], [71]. Future analysis could consider the broader social impacts of rooftop solar on electricity system costs and benefits. Additional costs include the rate impacts felt by households that do not generate solar energy, and any increase in the cost of electricity distribution system infras-tructure to enable solar self-generation. Additional benefits include household energy security, which may be possible if households can self-generate, store electricity, and “island” at times when power is lost on the grid.

We do not forecast declines in the cost of residential rooftop solar. Solar modules have dropped in price over the period of 2009-2021 [5] and future price drops would improve the financial returns from residential rooftop solar. Future work could evaluate the degree to which rooftop solar installed costs will decline as module prices fall and innovations are made to reduce the “soft costs” of resi-dential solar installations such as permitting and customer recruitment [72].

In previous research, Dolter and Boucher [33] outlined how the changes to SaskPower’s net metering program made in 2017 and 2018 emerged from a process of engage-ment with Saskatchewan residents and the solar industry. This process was an example of procedural energy justice, enabling a diversity of voices to contribute to program de-sign. The program changes SaskPower made in 2019 were carried out without consultation and resulted in anger. The rationale for the program changes was that net metering leads to cross-subsidization as non-solar customers pay higher electricity rates due to load defection by customers who install solar. Equity issues arise when high-income households receive generous payments for installing solar, and higher rates pose a burden for low-income households. Those who we deem as solar-eligible were more likely to have incomes greater than $60,000 compared to the sam-ple as a whole. Future work can address the distributional equity of solar energy program design, and issues of proce-dural energy justice.