ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846
-A A +A

Predicting Food Price Inflation through Online Prices in India

Ritwik Banerjee ( and Chetan Subramanian ( are with the Indian Institute of Management Bangalore, Bengaluru. Nished Singhal (nished. is director, NeenOpal Intelligent Solutions, Bengaluru.

Food and beverages price data from a leading online marketplace is used to compute an online price index. This index successfully tracks the offi cial consumer price index not only at the aggregate level, but also at the subgroup level. The results indicate that online prices can be a quick and effi cient source of price data, which can provide a cheap, but credible, signal of infl ationary pressures on a real-time basis.

The online e-commerce market provides us with a unique opportunity to construct price indices and measure inflation. The advantages of using online prices over price data collected through traditional means are many. One, data collection is quicker, cheaper, and more efficient than conventional methods. Two, data can be obtained much more frequently through special softwares, which have capabilities of crawling through a vast number of items in online marketplaces. And, three, the quantum of data is often large and the quality better, since it circumvents the possibilities of human error.

However, online prices may not have some other desirable characteristics. Even though the e-commerce industry is growing at the rate of over 60% per year, market penetration has largely taken place only in the 10-odd major metropolitan areas of India. Consequently, it is a challenge to replicate the basket used to calculate official statistics, using online prices only. Therefore, whether online prices track the official price index is an empirical question. This article is an attempt to address this issue.

While attempts have been made to creatively use online prices to predict inflation in Latin American and other countries (Cavallo 2012; Cavallo and Rigobon 2016), there is limited literature in this context in India. A notable exception is the attempt by Das et al (2016), who use online prices to track the direction, but not the intensity of prices of select food items from the official consumer price index (CPI) basket. By contrast, we construct a measure that allows us to create the appropriate counterfactual: what if the official CPI was constructed using the online data? Overall, this article aims to add to the research by validating online price data with official inflation statistics in India. We view this exercise as a necessary first step before online price data can be used for policy purposes.

In this article, we have focused only on food and beverages since a large share (45.86%) of food in India’s inflation means food prices affect inflation directly (Basu 2011; Rajan 2014) as well as through increase in non-food prices and rural wages (Guha and Tripathi 2014). Also, a very high correlation of 0.978 between the food and beverage inflation index and the overall CPI inflation (data from January 2013 to January 2016) indicates why tracking food and beverage prices is necessary for tracking overall inflation. Together with these, the relatively easy availability of food and beverage items in online marketplaces makes it a suitable category for us to track.

For the construction of the online index, we have used price data of items in the food and beverages category from one of the largest online retailers in India, namely Bigbasket, from April 2015 to July 2016. The data set contained detailed information on each product, including prices, product identity, and category indicators. We use official weights and standard methodologies to compute CPI on the online price data, to arrive at the online price index. Our results show that the online index successfully tracks both the direction and the magnitude of official CPI. This is more robust than alternative methodological specifications.

Data and Method

The data was obtained from Bigbasket, an online retailer based in Bengaluru, India. The data was restricted to items in the food and beverage category and prices were collected for the Bengaluru metropolitan area, which is the largest urban market in Karnataka and one of the largest in India. We restricted ourselves to Bengaluru despite online data being available from other metros in India due to a serious concern of missing data. The market share of the online retailer within the city’s grocery market is about 1% of the total and about 10% of the organised market.1 Despite the low market share, the e-retailer prices are a very good representation of the entire urban market of the city. The e-retailer tracks the prices of other modern trade retailers on a daily to weekly basis as well as the mandi prices of fruits and vegetables on a daily basis, and corrects its prices.

The categories in food and beverage groups of the official inflation basket are as follows: cereals and products, meat and fish, egg, milk and products, oils and fats, fruits, vegetables, pulses and products, sugar and confectionery, spices, prepared meals, snacks, sweets, etc, non-alcoholic beverages. For each of the items2 in the above categories, we obtained one or more corresponding online products from the retailers’ website. Those items for which prices were available for most of the 16-month duration were given preferences. Within cereals and products, for key items, such as rice and atta (wheat), more than one item was selected for two main reasons. One, price movements may vary for the same item across different varieties. Consequently, even the Karnataka urban CPI tracks the price of two varieties of rice. Two, tracking price of only one product makes it more susceptible to brand-specific price fluctuations.

The online data contains a combination of food and beverages. Prices of over 120 products across 25 categories, 12 subgroups and two groups were tracked daily from April 2015 to July 2016, that is, for a total of 16 months. The definitions of categories and subcategories and their respective weights are exactly the same as those in the official index. This helps ensure that there is no difference between the online inflation index and the CPI due to the choice of the items in the inflation basket or their weights. Thus, any differences will only be because of differences in the item-level price movements of the online and official indices. Table 1 describes the data.

Exceptions such as missing data have been handled in line with the procedure followed by the CPI.3 First, items with negligible weights are not tracked separately and the price movements of the subgroup are assigned to the item. Second, prices of seasonal items, such as mango, which were unavailable, have been substituted with price changes of the item with the largest share in the subgroup (banana for fresh fruits). Third, short gaps in individual price series are filled by carrying forward the last available price. Fourth, items such as rice, sugar, and atta are also distributed at subsidised prices through the public distribution system (PDS). However, they account for a negligible change in the inflation index as their weights as well as prices are very low. Hence, their prices have been assumed to be constant for the PDS component. And fifth, in rare scenarios such as that of apples in our case, where existing variety of items were discontinued during the tracking period and a new variety was introduced, a relative price comparison at the time of change was made and accordingly prices were adjusted in line with the methodology followed by the CPI to handle discontinuation of the tracked item.

Another challenge faced in constructing the online index was the difficulty in mapping cooked items, that is, finding online substitutes for items such as prepared meals and snacks, cup of tea, etc. The closest substitutes were items such as ready-to-eat meals. However, the price data was missing for most of the period and hence, the same was not used for tracking. Thus, for two subgroups—prepared meals, snacks, sweets, etc, and non-alcoholic beveragesthe price index from the CPI has been used directly in the online index.


In order to evaluate the performance of the online index based on its ability to track the official index, we first start with the comparison of the two indices at the aggregate level and attempt to explain the similarity as well as differences between the numbers generated by them. Figure 1 compares the online index developed for food and beverages with the CPI for Karnataka (urban food and beverages) for the period May 2015 to July 2016 (April 2015 index value is taken as base = 100 for both the indices for ease of comparison).

The close co-movement between the two indices is captured by the high correlation coefficient of 0.935 (p-value<0.01). This suggests that the online index tracks the CPI quite well. The mean of the online index is lower than the CPI; this is primarily driven by the sharp dip in the food inflation during February–March 2016. The online index shows a higher dip than the CPI in February–March 2016 and, thus, a lower mean inflation. This lower inflation could possibly be due to one or a combination of the following reasons: high responsiveness of the online retailer to drop in market prices, or product price changes by the retailer due to internal reasons (such as change in pricing strategy, discounts for customer acquisition, etc).

The graph in Figure 1 and high correlation values observed demonstrate the capability of online indices to track the official inflation index. Although, we can observe one period (February 2016 to April 2016) where there is some difference between the two indices, by and large the online index is able to track both the level and trend of the official index quite well.

In order to understand the reasons for the dip in the online index, we further analysed the month-in-month change of the online index at the subgroup level. The subgroup-level comparison between the online index and CPI will also help us understand the relationship between the two indices at a much more disaggregated level. Clearly, success depends not only on the ability of the online index to track CPI at the aggregate level but also at a more granular level. The correlation coefficients between the online index and the CPI for each subgroup are summarised in Table 2. Figure 2 gives a snapshot of the co-movement between the two indices for the top six subgroups.

Table 2 and Figure 2 indicate that a majority of the subgroups exhibit a close comovement with high correlation coefficient between online index and CPI. In fact, all the correlation coefficients are statistically significant at the 5% level. The panel in Figure 2 for the subgroup “vegetables” further suggests that the accentuated seasonal dip captured in Figure 1 is largely driven by this particular subgroup. No such dip is observed in any other subgroup category. Clearly, not only does the online index track the official CPI well for food and beverages, the online index is also successful in tracking the price index of all the major subgroups within food and beverages.


Median CPI

As we have observed in the data and the analysis, the price index for some of the items in the food and beverages basket are quite volatile. There are seasonal effects in subgroups such as vegetables. In this section, we use an alternative methodology to define a price index which can potentially decrease the index volatility without sacrificing its ability to track the substantive aspect of price movement and, thus, may be considered to be a more effective measure of inflation by some accounts. In particular, we employ the weighted median CPI methodology developed by Bryan and Pike (1991) and Bryan and Cecchetti (1994). Recently, Ball et al (2015) used WPI data to find that weighted median inflation in India is substantially less volatile than headline inflation and this is consistent with observed patterns in advanced economies.

Thus, instead of arriving at the food and beverage index using a weighted average of the index of the 12 subgroups, in this section, we calculate the weighted median index of the subgroups for each of the months covered in the analysis. To construct the weighted-median CPI for the food bundle, we calculate the percentage changes of all the subgroup prices in the CPI food bundle and then find the median value of the price changes. An unusually large increase in the price of vegetables or any other item will not change the median CPI unless it jumps from the group below the median to the group above the median. Even then, since it is just one of many prices in that group, it will not change the median very much. Only if most subgroup prices increase will the median increase noticeably. Therefore, such an approach neutralises volatile price changes.

Figure 3 plots the weighted median online price index with the official CPI for food and beverages. The weighted median CPI does not dip in February–March 2016 unlike the official CPI, consistent with the view that the weighted median CPI estimates the stable trends in inflation and ignores the short-term fluctuations. To the extent that this is a desirable property of a computed index, we can now analyse if the online index captures the essential trends in inflation by comparing the weighted median online index with web-based prices and the weighted median official CPI. This is plotted in Figure 4.


As Figure 4 indicates, the indices when computed with the weighted median methodology demonstrate a very close co-movement, sometimes even closer than what we see in Figure 1. The correlation coefficient between the weighted median online and weighted median CPI is 0.955 (as opposed to the correlation coefficient between the online index and official CPI of 0.935). The reason for the improvement in the quality of tracking is that the median methodology successfully dampens the effect of volatile components such as prices of vegetables in both the online as well as the official indices. Thus, if the stable inflation trends are of policy interest, then the online index is actually a good proxy for the official CPI.


We constructed an online price index using food and beverage data from a prominent online marketplace and the official weights used to calculate CPI in India. We show that the online index is capable of tracking the level as well as the dynamics of CPI. Results show that even at a more disaggregated level, despite some discrepancies, the online index does well in capturing the warp and the woof of the price movements of the product subgroups. Our article is not meant to be a definitive exercise on the validation of online prices, but is aimed at placing this idea as an important one in the context of policymaking in India. We hope more rigorous empirical exercises will be conducted to validate online price data in the future.


1 Estimates as per the e-retailer.

2 Source of Items & Weights is NSSO (2014).

3 The methodological details are reported in CSO Report (2014) and CSO Report (2015).


Ball, L, A Chari and P Mishra (2015): “Understanding Inflation in India,” NBER WP No w22948, National Bureau of Economic Research.

Basu, K (2011): “Understanding Inflation and Controlling It,” Economic & Political Weekly, Vol 46, No 41, pp 50–64.

Bryan, M F and C J Pike (1991): “Median Price Changes: An Alternative Approach to Measuring Current Monetary Inflation,” Economic Commentary, Federal Reserve Bank of Cleveland, 12 January.

Bryan, M F and S G Cecchetti (1994): “Measuring Core Inflation,” Monetary Policy, N Gregory Mankiw (ed), Chicago: University of Chicago Press, pp 195–215.

Cavallo, A (2012): “Online and Official Price Indexes: Measuring Argentina’s Inflation,” Journal of Monetary Economics, Vol 60, No 2, pp 152–65.

Cavallo, A and R Rigobon (2016): “The Billion Prices Project: Using Online Prices for Measurement and Research,” Journal of Economic Perspectives, Vol 30, No 2, pp 151–78.

Central Statistics Office Report (2014): “Report of the Group of Technical Advisory Committee on Statistics of Prices and Cost of Living,” Prices & Cost of Living Unit, National Accounts Division,

— (2015): “Consumer Price Index, Changes in the Revised Series,”

Das, A, P Ghosh and A Sanyal (2016): “Monitoring Consumer Price Trend Using Daily Price Data of Online Grocery Stores in India,” United Nations Asia-Pacific Economic Statistics Week, Seminar Component, Bangkok.

Guha, A and A Tripathi (2014): “Link between Food Price Inflation and Rural Wage Dynamics,” Economic & Political Weekly, Vol 49, Nos 26–27, pp 66–73.

NSSO Report (2014): “Household Consumption of Various Goods and Services in India 2011–12,” Report no 558, NSS 68th Round,

Rajan, Raghuram (2014): “Fighting Inflation,” inaugural speech at FIMMDA-PDAI Annual Conference 2014, Mumbai, 26 February.

Updated On : 8th Jun, 2018


(-) Hide

EPW looks forward to your comments. Please note that comments are moderated as per our comments policy. They may take some time to appear. A comment, if suitable, may be selected for publication in the Letters pages of EPW.

Back to Top