• No se han encontrado resultados

4. Aplicación De La Metodología Para El Análisis De Los Barrios

4.2 Delimitación de los siete barrios de estudio

We use data from several administrative sources. The main data are more than 117 million daily

observations of natural gas consumption from June 1, 2015, to July 1, 2017. Based on discus-

sions with SoCalGas, we exclude a small number of extreme outliers. We do not use negative

reads (N =64) or reads above 71.8 therms per day (N =526; the 99th percentile of daily con- sumption is 7.18 therms). In addition to consumption data, we have access to monthly billing

data from January 2014 to June 2017. For each bill, we observe the total consumption and monthly

utility charge.

Panel A in Table 1 provides summary statistics for the pre-experiment period. Between June

1 and October 31, 2015, the average customer used about 0.7 therms per day (with a standard de-

viation,σ, of 0.6).21 From June 2014 to October 2015, average monthly gas usage was about 39

therms (σ=25.8), and the average utility bill was $48 (σ=31.7).22 The total amount of natural

19SoCalGas excluded low-use customers from information treatments based on insights from the first and sec-

ond year of the conservation campaign and previous findings in the literature showing no responses to non-price interventions by customers in the lowest usage deciles (Allcott, 2011c; Ferraro and Price, 2013).

20MyAccount is SoCalGas’s customer online portal. It provides bill overviews, historical usage feedback, bill

payment capabilities, and information about ways to save energy. Another reason for requiring MyAccount partic- ipation was that the regulator mandated that information collected through smart meters had to be made available to all customers. Fulfilling this mandate without obstructing a clean measurement of the impact of the conservation campaign required a baseline intervention. All customers, including the control groups, were invited to use the My- Account portal after installation of the smart meter. Our estimated treatment effects are always net of effects that may be caused by the access to MyAccount.

211 therm describes a unit of heat energy equal to 100,000 British thermal units (BTU).

22Customer are billed for three elements: (i) procurement costs based on forecasts of wholesale market conditions;

(ii) transportation costs that cover gas delivery and infrastructure investments; and (iii) public purpose surcharges, such as fees to fund the California Alternate Rates for Energy program or R&D expenses.

Table 1: Pre-Experiment Summary Statistics and Treatment Balance

All Control BTA Difference

(N=154,202) (N =74,202) (N=80,000) (µBTA−µC)

Panel A: Usage Behavior and Billing

Daily Usage 0.7017 0.7014 0.7019 0.0005 (0.5989) (0.6022) (0.5959) (0.88) Monthly Usage 38.9249 38.9545 38.8974 -0.0571 (25.7722) (26.1115) (25.4535) (0.66) Total Bill 47.8798 47.9026 47.8586 -0.0440 (31.7329) (32.0809) (31.4068) (0.79)

Panel B: Account Details

Billing Factor 1.0553 1.0552 1.0554 0.0002 (0.0138) (0.0133) (0.0143) (0.01) 1(Level Pay) 0.0396 0.0401 0.0391 -0.0010 (0.1950) (0.1963) (0.1937) (0.28) 1(Paperless Billing) 0.6689 0.6688 0.6690 0.0002 (0.4706) (0.4706) (0.4706) (0.94) 1(Account Closure) 0.1449 0.1440 0.1457 0.0017 (0.3520) (0.3511) (0.3528) (0.34)

Account Tenure (Years) 13.2862 13.3157 13.2587 -0.0570

(10.7344) (10.7558) (10.7144) (0.30)

Panel C: Property Characteristics

1(Individually Metered) 0.9887 0.9887 0.9886 -0.0001

(0.1059) (0.1055) (0.1062) (0.77)

1(Single Dwelling) 0.8485 0.8493 0.8477 -0.0016

(0.3586) (0.3578) (0.3593) (0.39)

1(Property in Climate Zone 1) 0.9478 0.9481 0.9476 -0.0005

(0.2224) (0.2219) (0.2228) (0.70)

Notes:Summary statistics for the full sample and by treatment group in the pre-experiment period (June to October 2015). We report means and the corresponding standard deviations in parantheses. Column 2 reports averages and standard deviations for all customers; column 3 and 4 restrict the sample to customers in the control and pooled treatment group, respectively. Column 5 presents differences in raw means between the control and pooled BTA group, and the p-values of two-samplet-Tests in parentheses. Table A1 in the appendix reports the same information by BTA treatment subgroup.

gas delivered to a customer is adjusted by a billing factor to reflect the heat value of gas at differ-

ent altitudes. SoCalGas uses two pricing tiers for residential customers: (i) a baseline allowance

with a low marginal price; and (ii) consumption above the allowance at a higher marginal price.

We also obtained customer-specific data, such as the treatment assignment, account activation

and deactivation dates, and indicators for participation in other utility programs. Panel B in Table

1 reports the corresponding averages and standard deviations. Around four percent of customers

participate in the Level Pay Plan, which is a payment plan that distributes annual energy charges

equally over twelve months. Approximately two-thirds of customers in our sample receive pa-

perless bills in June 2017. In total, about 15 percent of customers close their account during our

sample window or around 5 percent per year.23 Most accounts were closed because the main ac-

count holder moved out of the home. The average customer has received natural gas service from

SoCalGas for about 13 years.

The third data type at our disposal is limited information about each property. This informa-

tion is summarized in panel C of Table 1. Almost 99 percent of all customers live in homes that

are individually metered. The remaining one percent live in properties that are centrally metered,

such as trailer parks, condominiums, or apartments with a master meter. About 85 percent of cus-

tomers live in detached single-family homes, about 7.5 percent live on properties with two to four

connected dwellings, and the remaining 7.5 percent inhabit a home with five or more connected

dwellings, such as apartment complexes. SoCalGas divides its service area into three different

climate zones. Close to 95 percent of the customers in our sample live in climate zone 1, about 5

percent live in climate zone 2, and 38 customers live in climate zone 3.24

Lastly, we collected daily weather data from January 2014 to June 2017 via NOAA’s Climate

Data Online search tool.25 Because most of our sample is from the Los Angeles metropolitan

23We present main results after exclusion of all customers who deactivate their account during the sample period

in Table A3 in the appendix. Account closure rates match similar programs in the literature (Allcott, 2011c; Brandon et al., 2017).

24SoCalGas defines the three climate zones as follows. Zone 1: Basin, Valley, South Coast, Low and High

Deserts; zone 2: Central Coast and the San Joaquin Valley; zone 3: High Elevation Mountains.

area, we use the average temperature reported by the Los Angeles International Airport (LAX)

weather station. Therefore, the data only vary over time but not across households.

Documento similar