The usefulness of the CMIE household survey data for electricity research in India

by Susan Das, Renuka Sane and Ajay Shah.

Measurement for electricity research

The problems of the electricity system are of particular importance in India given that the existing institutional arrangements work relatively poorly. While this has been a concern for decades given the importance of electricity in economic growth, it has achieved a fresh prioritisation due to the need for a clean energy transition. The problems of electricity sector have become the critical bottleneck for the decarbonisation of the economy (Jaitly and Shah, 2021).

De jure government subsidies, and de facto theft, take place in this field on a significant scale. While researchers are able to readily observe de juresubsidies, theft is hard to observe. Domestic consumers and agriculturists shape the political economy of sub-national electricity. Policy makers make decisions about the de jure and de facto policy frameworks with an eye on how these pressure groups will react. The mechanisms through which transfers take place are often subtle. As an example, Mahadevan (2023) obtained micro data for one state in India, and found that electricity bills were manipulated to charge less to households in regions that had voted for the ruling party.

Understanding the field, and analysing possible policy pathways, will benefit from quantitative political economy research grounded in household survey data. Once household level electricity expenditures are observed alongside an array of household characteristics, it becomes possible to analyse the incidence of present or alternative subsidy mechanisms. Researchers could try to understand the mechanisms through which tariffs and subsidies shape expenditures, the role of appliance ownership, the problem of theft of electricity, the elasticity of residential electricity demand to price changes and other policy actions, and the impact of electricity consumption upon the household.

These possibilities have been opened up in India through the CMIE CPHS database, a panel dataset where 240,000 households are observed thrice a year. As with many measurement strategies such as crime victimisation surveys, there is value in looking beyond aggregative measurement, and directly asking households about their experiences. The CMIE CPHS dataset opens up diverse array of research possibilities in the field of electricity research. Some research papers have emerged which harness this. Martínez Arranz et al. (2021) have used the CPHS to study the expansion of access to 24 hours of electricity supply between the 2014-2019 period across 24 states and UTs. Kulkarni, Sahasrabudhe and Chunekar (2022) have used the CPHS to analyse the appliance ownership trends in India which potentially can be used to forecast residential electricity demand.

At the same time, there are concerns about the veracity of this data. The world over, it is difficult to supervise field investigators and obtain cooperation from respondents, particularly from high income households. While the CMIE CPHS database represents a remarkable new phase in economic measurement in India, there has been a debate about the difficulties of this data.

In this article, we perform sanity checks of this data, across natural experiments of tariff changes in Tamil Nadu. This helps assess the extent to which the data can be a sound foundation for researchers in this field. It also contributes to the larger literature on the problems of household survey data, on economic measurement in India, and on pathways to measurement in emerging markets that do not rely on the official statistical system.

The CMIE CPHS database

The Consumer Pyramids Household Survey (CPHS) is a longitudinal household survey conducted by the Centre for Monitoring Indian Economy (CMIE) from 2014. A panel of over 240,000 Indian households across 27 states and 514 districts is measured thrice a year. The sample is nationally representative and selected through a multi-stage stratified design. The survey captures many features of the household including labour supply, investment, borrowing, consumption, income, and demographics. It represents the first large scale longitudinal dataset of this nature in India.

A rich literature has emerged based on the CMIE CPHS data, ranging from questions of household portfolio choice, labour markets to health and mortality impacts of diseases. In May 2024, a search for the string CMIE CPHS on Google Scholar showed 310 papers. Pandey, Patnaik and Sane (2019) study the evolution of financial savings with the changes in the design of tax breaks, while Gopalakrishnan, Ritadhi and Tomar (2019) study the effects of adjustment costs in real estate affecting portfolio choice. Patnaik, Sane and Shah (2019) study the effects of a flood in Chennai on the income and consumption pattern of affected households and the heterogeneity in impact by financial constraints and income levels. It has been used to understand gender gaps in the labour market outcomes as a result of the impact of COVID-19 (Deshpande, 2022; Abraham, Basole and Kesar, 2022, excess mortality and prevalence of COVID-19 (Malani and Ramachandran, 2022; Mohanan et al., 2021), and the impact of location and income on the health of individuals (Patnaik, Sane, Shah and Subramanian, 2023).

Household survey measurement is facing challenges the world over. Respondents who are immersed in modern distractions, or are busy, are prone to refuse to answer questions, which gives non-random non-response. Supervision and control of field investigators is difficult in the Indian locale, which makes field research particularly difficult. Publication orientation creates little incentive for researchers to expend resources on correctness of data (or software).

In the specific context of the CMIE CPHS database, there have been concerns about a potential under-counting of poor households. For example, Somanchi (2021) argues that CPHS under-represents women, young children and poor households and, over-represents well-educated households. Similarly, Pais and Rawal (2021) raise questions regarding the lack of a precise sampling frame and the likely exclusion of poor and mobile households. CMIE has responded to these concerns (Vyas, 2021). These debates raise concerns about potential applications of this database.

The purpose of this article is to examine the usefulness of one measure from this database – expenditure on electricity — where veracity of measurement is essential for applications in the field of electricity. Towards this objective, we harness a group of natural experiments, and examine the gross regularities seen in the data.

Natural experiments with the price of electricity in Tamil Nadu

We examine four events where the price of electricity changed in the state of Tamil Nadu. Two of these were tariff changes by the regulator (the Tamil Nadu Electricity Regulatory Commission (TNERC)), and the other two were changes in the on-budget subsidy by the state government.

Prior to the first event, the tariff regime prevalent in the state was based on the tariff order passed in 2012 by the TNERC. The effective prices for electricity varied between INR 3.00 – INR 3.25 per kWh for households whose consumption was below 100 kWh of electricity. Further, the prices gradually increased to the range of INR 3.50 – INR 4.60 per per kWh for households whose consumption was up to 250 kWh in a month. For consumption beyond 250 kWh, the price rose to INR 6.60 per kWh. In this backdrop, a sequence of four natural experiments took place:

Subsidy increase in 2016

During the state legislative assembly election campaign in Tamil Nadu in 2016, the then Chief Minister Jayalalithaa announced 50 kWh of free electricity a month to over 19 million residential consumers as one of her poll promises (Srikanth, 2016). She won the election, and the promise was kept in June 2016.

Tariff decrease in 2017

On 11 August 2017, the TNERC issued a tariff order which decreased tariffs for domestic consumers (TNERC, 2017).

Subsidy increase in 2018

In November 2018, the Government of Tamil Nadu (GoTN) introduced additional subsidies. These varied from INR 0.5 – INR 1.00 per kWh for consumption below 250 kWh in a month (TNERC, 2018). This is a small event where very little changed.

Tariff increase along with subsidy in 2022

The TNERC approved a tariff increase for domestic consumers across all consumption slabs in September 2022 (TNERC, 2022). Alongside this, the GoTN announced an additional 50% subsidy on electricity consumption between 50-100 kWh (for consumers having consumption below 250 kWh) in addition to the 50 kWh of free electricity (Guruvanmikanathan, 2022).

These changes are summarised in Table 1, which depicts the full price schedule of electricity vis-a-vis electricity consumption (kWh) for residential consumers in Tamil Nadu under different price regimes. We show the effective price factoring in the subsidies by the GoTN for a household as per its electricity consumption after each event of tariff or subsidy change announced by the state.

What might we expect to see, in expenditure data at a per-household level, across these four events? For the purpose of this article, we bring a simplistic prior: that the short-term price elasticity of energy demand is low, so the household expenditure will just go from qp₁ to about qp₂, with adjustments in the consumption basket based on a commensurate income effect only.

We recognise, of course, that the reality is more complex than this. Many other elements of adjustment are in fray, both in the short term and in the long run. With a lag, higher electricity prices change the incentives for energy efficient equipment and rooftop solar generation. Going beyond such conventional economic responses, the incentives for theft change: when electricity is more expensive there is a greater incentive to steal. The changes in the price schedule seen in Table 1 could induce complex effects combining these factors. Sophisticated research projects, in the field of household electricity consumption, are required that seek to tease out the short-term and long-term effects.

This article is not in that field: it is an examination of the concerns around measurement of household level electricity expenditure as seen in the CMIE CPHS database. We examine the gross regularities about how household electricity expenditure changed across these natural experiments. Our ability to do this across four natural experiments, adds up to an opportunity to examine the usefulness of the CMIE household survey data for the purpose of electricity research. If the expenditure change generally goes with the price change, for relevant households, we will conclude the dataset has value for researchers in this field.

Data description

We focus on the 10,000 households observed in Tamil Nadu. Within this, we obtain three facts:

• Total expenditure (INR) – This is sum of all the expenses incurred by a household on the consumption of goods and services during a month. It includes expenses on rent, food, clothes, utility bills, entertainment, fuel etc. This feature is not directly captured during CPHS survey process, rather it is derived as the sum of all the monthly individual expense heads captured during the survey of a household. We derive this feature from the monthly expenses data.
• Electricity expenditure (INR) – This is the expenditure incurred by a household on electricity during a month. We obtain this feature from the monthly expenses data.
• Weights – We use household weights for the state level estimates provided by CPHS for making estimates at the monthly frequency level. We further use the adjustment factor for household non-response to adjust sampling weights to balance for non-responses during the survey. It is the ratio of the total number of sample households in the stratum and the non-surveyed regions to the accepted sample from these.

An important strand of the field of economic measurement is assessing how a variety of methods for measurement fared through the pandemic. In this article, however, the third event was in November 2018 (which was well before the pandemic) and the fourth event was in September 2022 (which represents post-pandemic conditions). This article, thus, does not offer insights on data quality in the pandemic.

What might we expect?

1. The event: In June 2016, the state government announced 50 kWh of free electricity to domestic consumers, an implementation of a promise that was made in the election campaign.

   Prediction: We expect to see lower expenditure in the aggregate.

2. The event: The TNERC approved a tariff cut in its 2017 tariff order for domestic consumers with consumption below 250 kWh of electricity.

   Prediction: We expect to see lower expenditure in the aggregate due to tariff cut in addition to existing free 50 kWh of electricity.

3. The event: In November 2018, the state government announced a further subsidy on electricity sold to domestic consumers. Varying subsidies were introduced ranging from INR 0.50 – INR 1.50 per kWh for different slabs of consumption of up to 250 kWh of electricity consumption. This was in addition to the already existing subsidy of 50 kWh of free electricity across all slabs of consumption. This was the smallest of the four changes in the price schedule that are examined in this article.

   Prediction: We expect to see lower expenditure in the aggregate, with a small effect.

4. The event: The TNERC approved a tariff increase in its 2022 tariff order for domestic consumers. Alongside this increase in tariffs, the subsidy given by the state government also increased. GoTN announced a further 50% subsidy on the next 50 kWh of electricity consumption (for households below 250 kWh of consumption) in addition to the prior 50 kWh of free electricity.

   Prediction: We expect to see higher expenditure in the aggregate in smaller magnitude due to combination of tariff hike by TNERC and subsidy announcement by GoTN.

Results

We now examine the household survey data and assess the extent to which the predictions, based on a simple prior, are borne out by the statistical evidence. For each price change, we examine household expenditure in the six months before and after the event. We view this in the aggregate. The results are shown in Table 2.

Comments

Click here to cancel reply.

Your Comments

MOST RECENT

Load more ...