Using High Frequency Electricity Data to look at Economic Activity Impact of Coronavirus Lockdown

Leave a reply

Following a national address by PM Narendra Modi on 24th March 2020, a nationwide lockdown to contain the spread of the coronavirus was initiated. This lockdown involved halting of all economic activities barring those that were need to maintain essential supplies. Due to earlier confusion as to the details, even essential supplies took a hit. People were confined to their residences, and movement were restricted. All major service organizations shifted to a work from home mode of carrying out business.

Following the first round of lockdown, between 24th March to 14th April, a second continuation was declared starting 15th April to last till 3rd May with conditional relaxations in some regions where spread had till then been contained. The last round of lockdowns with further easing were continued from 3rd May till 17th May first and then later from 17th to 31st May by the National Disaster Management Authority. Following the lapsing of lockdown on 31st May, an Unlock 1.0 phase was declared starting 1st June with phase wise opening of all major industrial and retail activity.

Our data of daily consumption of electricity comes from the National Load Despatch Centre of the Power System Operation Corporation Limited which optimum scheduling and despatch of electricity through the national grid. They release daily, weekly and monthly reports of power consumed, energy supplied being reported in million units of energy supplied to each state grid. Using this data, we can then track the change in consumption of electricity nationwide which becomes a proxy for economic activity following from Cicala (2020) and Benedikt and Radulescu (2020). The fall in consumption of electricity can be a predictor of total output contraction in the period and the recovery in consumption can be a harbinger of the expected speed of recovery.

We remove seasonal components from the data that can distort our results, these seasonal aspects come from whether a time period is the harvesting period or are summer months which can impact electricity consumption. [i]

Overview of Power Consumption in India

India has been seeing a steady increase in Electricity consumed with a trend year on year growth rate of 3.5% with significant heterogeneity amongst state wise consumption growth rates. Bihar leads the electricity consumption growth rate with growth rate around 13% with North East states following close with rates ranging from 7% to about 10%. Amongst larger states, Telangana, Madhya Pradesh and Uttar Pradesh show higher rates of growth.

Figure 1 represents the per capita[ii] mean annual power consumption.

Figure 2 presents the overall trendline for India till 2019.

Impact of Coronavirus Lockdown

To estimate the impact of the lockdown we use data from financial year 2014 onwards to estimate the expected consumption during the lockdown period, accounting for the seasonality and trend. Using those trends and seasonality, we find that beginning the lockdown period there was a significant deviation between the actual and the predicted consumption of electricity. The actual consumption of electricity nationwide was consistently lower than the predicted one across all phases of the lockdown. This fall from the predicted value was the most for the 1st period of lockdown with an average fall of 12.25%. This shortfall in electricity consumption decreased across the phases with the 2nd period of lockdown witnessing only an 8% decrease while the further phases had no statistically significant difference from similar periods in previous years.

Table 1

Phase of Lockdown Change in Consumption compared to same period in previous years
Phase 1 (25th March – 14th April) -12.25%
Phase 2 (15th April – 3rd May) -8.07%
Phase 3 (4th May- 17th May) No statistically significant difference
Phase 4 (17th May – 31st May) No statistically significant difference
Unlock 1.0 (1st June onwards) No statistically significant difference

Further, as we can see from Figure 3 since the beginning of lockdown 3.0, we have seen an increase in actual consumption of electricity which is consistent with the increasing economic activity as restrictions were eased across rounds of lockdown.

Figure 3 The Actual and Predicted Consumption across the various phases of Lockdown

We also note an increase in consumption of electricity across the phases with phase 2 seeing a 5.4% increase in consumption. Table 2 provides the average increase in electricity consumption over the previous phase of lockdown. We see that each round sees an increase over the previous round till we come to Unlock 1.0 which saw no significant difference as compared to phase 4 (at 5% level).

Table 2

Phase of Lockdown Change in Consumption compared to the previous period of lockdown
Phase 2 (15th April – 3rd May) 5.38%
Phase 3 (4th May- 17th May) 8.60%
Phase 4 (17th May – 31st May) 10.55%
Unlock 1.0 (1st June onwards) No statistically significant difference

We thus see the data showing up the kind of trends we expect to see. The national level statistics do hide considerable heterogeneity in state level consumption of electricity with certain states exhibiting above average decreases while some states surprisingly see an increase in consumption. For more on this, stay tuned for the 2nd part in this series.

References

  • Cicala, Steve. Early Economic Impacts of COVID-19 in Europe: A View from the Grid. Tech. rep. Online, last accessed: May 6, 2020. University of Chicago, 2020.
  • Janzen, Benedikt, and Doina Radulescu. “Electricity Use as a Real Time Indicator of the Economic Burden of the COVID-19-Related Lockdown: Evidence from Switzerland.” (2020).

[i] Our analysis excludes Andaman and Nicobar Islands and Lakshadweep islands as they are not connected to the National Grid

[ii] Per 2011 population census

About the Author:

Ashutosh Dwivedi is a Research Associate at SRITNE.

Follow us on twitter: #ISBSRITNE

Follow us on LinkedIn: https://www.linkedin.com/showcase/sritne

Positive Spillover Effects of Uber on NYC’s Yellow Taxis

Leave a reply

The entry of for-hire vehicles (FHV) through on-demand taxi service companies such as Uber and Lyft have disrupted the traditional taxi service industry. The unique business model of these companies allows cab drivers to ply on the road without the need for the coveted million-dollar taxi medallions. Thus, the heavily regulated taxi industry saw a rise in the number of taxis on the road. This increase in the supply of taxis resulted in demand for rides shifting from the incumbent taxis, such as the Yellow taxis of New York, to the new entrants in the market, Uber, Lyft, and the like. Since the Yellow taxis depend on their daily trips for their earnings, a fall in the number of trips will consequently result in a fall in their earnings. Thus, leading to the general belief that the entry of the new entrants resulted in a decline in earnings of the incumbent taxis. In this study, we show that the presence of these new on-demand taxi service companies does disrupt the traditional taxi market by taking away the demand from incumbent taxi drivers. However, the incumbent taxi drivers can learn on the job and hence can increase their overall total earnings despite the fall in demand.

This study uses trip level data of Yellow taxis and FHVs (comprising of Uber, Lyft, and the like) in New York City for 2015-2019. We use this trip level data to compute Yellow taxi driver’s earnings, trip count, and the distance traveled. We calculate the trip count of FHVs, and we use this to measure the intensity of FHV’s presence in a given taxi zone and time. Using OLS, we compute the effect of FHV’s presence on Yellow taxi driver’s earnings, distance traveled, and trip count.

The number of Yellow taxis plying on the roads of New York City at a time is fixed by the number of medallions, which is, in turn, regulated by the Taxi and Limousine Commission. Most of the Yellow taxi drivers lease a taxi daily. This encourages drivers to set a daily wage target, and upon reaching that target, they go off the streets. This form of labor supply discourages taxi drivers from maximizing their earnings and utility by working more on high paying days and less on low paying days. The entry of additional taxis in the form of Uber, Lyft, and the like increased the number of cabs, thereby changing the dynamics of the market by increasing the supply of taxis. Hence, we expect the demand would now be spread across the increasing supply base. This translates to reduced demand for an individual Yellow taxi, and one would expect that they will be unable to meet their daily earnings targets or be forced to reduce their daily earnings targets. Eventually, resulting in a fall in their earnings per trip.

Our results support the initial hypothesis of reduced earnings. An increase in FHV trips by 24 rides result in a fall of Yellow taxi’s average hourly fare per trip by 8 dollars, and a reduction in Yellow taxi’s average distance traveled per trip by 20%. These results are for the entire city of New York. However, these results are not the same when we consider parts of the city at a time. That is, the effect of FHV’s presence differs by taxi zones. 1 In taxi zones, where traditionally (before Uber’s entry), you have high Yellow taxi activity (demand), we notice a fall in demand for Yellow taxi with an increase in FHV’s activity (FHV trip count). That is, with an increase in FHV trips by 24 rides, the Yellow taxi’s average hourly fare per trip falls by 10 dollars. While the average distance traveled per trip falls by 13%. These results hint towards a reduction in the supply of Yellow taxis with an increase in the supply of FHVs. That is a substitution between the supply of Yellow taxis and FHVs.

When we consider taxi zones where traditionally you have had low Yellow taxi activity, the supply of Yellow taxi rises with the same increase in FHV’s activity. An increase in FHV trips by 24 rides will result in a reduction of Yellow taxi’s average hourly fare per trip by 49 dollars. However, in these taxi zones, the average distance traveled per trip increases by over 220%. Here, the results point towards an increase in the supply of Yellow taxis with an increase in the supply of FHVs. That is, the supply of Yellow taxis and FHV are complementary to each other.

Manhattan and Brooklyn are two boroughs where we, traditionally, notice high Yellow taxi activity. This can be seen through the heatmap below. Thus, we re-run our results by comparing Manhattan and Brooklyn against the other three boroughs: Bronx, Queens, and Staten Island. These results confirm our previous findings; a substitution between the supply of Yellow taxis and FHVs in Manhattan and Brooklyn and a complementary effect between the supply of Yellow taxis and FHVs in the other three boroughs.

So far, we have looked at the average hourly earnings per trip and distance traveled per trip for an individual taxi driver. However, another crucial element of a taxi driver’s total earnings is the number of trips it makes in a shift. When we account for the number of trips made, we are able to calculate the overall earnings and total distance traveled by all taxi drivers during a given shift. These aggregate results indicate that the total hourly earnings and total distance traveled increases by 1348 dollars and 116 miles, respectively, with a concurrent rise in FHV trips by 24 rides. We are thereby indicating that overall, the presence of Uber, Lyft, and the like increases the demand for Yellow taxis.

Figure 1 Heatmap depicting Yellow taxi trips by taxi zone before Uber’s entry. A darker shade indicates a higher number of trip

A heightened FHV presence leads to a fall in earnings and distance traveled for an individual taxi driver. However, this individual-level effect of FHV is heterogeneous as Yellow taxi drivers are learning on the job and moving towards areas where they face less competition by the new entrants and increased demand. Further, upon accounting for the number of trips made by Yellow taxi drivers, we find that the total earnings and distance traveled of all taxi drivers increases with an increase in FHV trips.

We suggest that this phenomenon can be explained by a behavioral shift that the presence of Uber, Lyft, and the like have brought about. We postulate that the people are now habituated to taking taxi rides for commutes and hence, increasing the total demand for taxi rides. The increase in demand must be significant enough for it to spill over onto Yellow taxis despite the rise in taxi supply. Testing of this theory is beyond the scope of this study, and we leave it open for future studies to explore.

Previous studies have documented a reduction in the total earnings of the incumbent taxi driver; therefore, this study is contrary to these previous studies. Though promising, we have only used New York city taxi trip data to arrive at the results. Hence, they are not indicative of a similar phenomenon in other cities/countries. Due to data limitations, we could not calculate the distance or the fare for FHVs, nor could we estimate our results at an individual driver level. With an increase in the availability of high-resolution data covering more geographic regions, we will be able to answer these questions for more cities and with better precision.

 

About the Author:

Rohin Nandakumar is a Research Associate at SRITNE.

Follow us on twitter: #ISBSRITNE

Follow us on LinkedIn: https://www.linkedin.com/showcase/sritne

 

Women In Technology And Innovation

Leave a reply

The early pioneers of technology and innovation were woman who, for the most part of  history, were unheralded. Marie Curie, Ada Lovelace, Grace Hopper, Anita Borg, among many others, have worked on groundbreaking inventions and computations in the field of nuclear physics, human-computer interaction, astronomical crunching, programming language compilations, etc. Only now these contributions are being recognised and rightly attributed. It is fair to say that woman participation, by and large, is being encouraged and supported: with organisations taking conscious efforts for inclusion, with measures undertaken to address issues of pay parity, woman’s return to work, education for girls, etc. But this improvement is rather trivial. The tech community largely reeks of gender- (and minority) based inequalities in terms of opportunities and pay. The proportion of female executives considerably dwindles along the hierarchical ladder, and even more so in the STEM community. Gender inequality in technology and innovation is a stigma that calls for immediate action.

Ongoing research at the Srini Raju Centre for IT and Networked Economy examines the biases against woman in technological innovation. The research was primarily motivated by the need to discern the under-representation of minorities, specifically, woman in patenting. The results highlight the presence of significant gender-based bias[1] in obtaining patents.[2] We create a measure of novelty in patent application and discuss the consequences of producing different technology for female over male inventors. Uniqueness or novelty of a patent application is calculated as its inverse similarity to those filed in the previous year in the same technological class. We observe that higher dissimilarity is rewarded with a higher rate of acceptance, higher maintainability, more citations, etc. Contrariwise, dissimilar patents with a high proportion of female inventors do not see equivalent acceptance.

Our findings show that woman inventors face higher penalties – lower rates of patent acceptances, more days to grants, higher efforts for maintenance – for novelty. These costs hold true for patent applications in varied scenarios, such as:

  • Technological classes like “Computers and Communications”
  • Single woman inventors
  • All female inventors
  • Women in large organisations
  • Woman-dominant patent applications (proportion of female inventors is above 50 percent)

Below visualisations were developed from the data developed at SRITNE and I attempt to visualise the findings. Figure 1 illustrates the average number of male and female inventors for patent applications over years from 2002 to 2014. A significant increase in the number of inventors is observed, while the number of female inventors is constant. It is important to highlight the gap in patenting by women and men.

Figure 1: Average number of inventors by gender over time

Figure 2 represents the average proportion of female inventors by year for different patent technological categories: Chemical, Food, Electrical & Electronic, Textiles, Mechanical, Others, Computers & Communications and Drugs & Medical, as defined by NBER.

Figure 2: Number of applications filed vs Number of Patent Issued by all

Figure 3 shown below illustrates the probability patent acceptances for different team sizes as the proportion of female inventors for the given team increases. We see a reduction in the rate of patent issued as the proportion of female on an application increases.

Figure 3: Probability of patent being issued for different team sizes, with number of inventors = 1, 2, 3, 4, 5, 6

Figure 4 below shows the average total citations received by applications for different team sizes as the proportion of female inventors for the given team increases. We see a reduction in the total citations as the proportion of female on an application increases.

Figure 4 demonstrates the total citations received by patents for different team sizes with increasing female proportion.

Clearly there are biases against woman in technological innovation. These results have been supported by multiple gender-based studies. A lot of research has reiterated that women at an organisation’s innovation frontier can lead to improved business performance. Thus, their under-representation and lack of opportunities in the workforce subsumes a prejudice rooted in the society. Women, single handedly, have come a long way from disenfranchisement to multiple waves of feminism. It is only right that the entire community wholly devotes itself to creating a gender-equal workforce.

1 Gender identification has been done using the API provide by https://genderize.io

2 Only patents filed and granted at the United States Patent Office (USPTO) have been used for the study

 

About the Author:

Bhoomi Thakkar is a Research Associate at SRITNE.

Follow us on twitter: #ISBSRITNE

Follow us on LinkedIn: https://www.linkedin.com/showcase/sritne