comments 4

Has the Indian demonetization served its primary purpose?

Disclaimer

This is probably a sensitive (and hopefully not so controversial) post- because it involves more than a few ‘already’ sensitive issues like terrorism and demonetization. So I start with some disclaimers.

  1. This is the work of a full time employed, wife loving, cricket playing, cricket watching, movie binging, petarap liking, food loving bloke, trying to have fun with data in the time spared after all the aforementioned activities.
  2. I trust the news channels to do their job properly. (i.e assume that if there is a terrorist attack in city X, with Y casualties and Z injuries, the channels do not fail to report that). All I do is get (scrape) the news article from them. Should you have any issues with the news content, feel free to reach out to the respective channel.
  3. I do not have any political affiliations and remain fiercely neutral. But it doesn’t mean I’m not critical of policies of either leanings.
  4. A lot of the quotes are from Wiki

And now we begin.

On the 8th of November 2016 at 20:00 Indian Standard Time, Government of India, headed by the Prime Minister Mr. Narendra Modi appeared on a unscheduled live televised address to the nation, announcing the immediate (taking effect in less than 4 hours) ‘ban’ of the usage/validity of all ₹500 and ₹1,000 banknotes of the Mahatma Gandhi Series as legal tender. This has since been called Demonetization, followed by some or many, pro or anti sentiments (a la, the demonetization fiasco) depending on who you ask or what news channel/paper you subscribe to.

The government claimed that the demonetization was an effort to stop counterfeiting of the current banknotes allegedly used for funding terrorism, as well as a crackdown on black money in the country. The move was also described as an effort to reduce corruption, the use of drugs, and smuggling.

Reactions

The defiantly neutral yet patriotic author, has been closely following, albeit through the eyes of the media. The ‘demonetization is good’ folks comprises of India Inc., featuring big names like Arundhati Bhattacharya (Chairperson of SBI) and Chanda Kochhar (CEO of ICICI), Anand Mahindra (Mahindra Group), Sajjan Jindal (JSW Group), Kunal Bahl (Snapdeal), some of them claiming that it would also accelerate e-commerce, etc.. Infosys founder N. R. Narayana Murthy has also praised the move. “These days, they don’t frequently see transportation of children by traffickers in these areas and they are not able to operate so easily because they cannot use the black money which funded their illicit businesses before,” said Kailash Satyarthi (Child rights activist and Nobel Laureate) to the Thomson Reuters Foundation.

On the other side, Nobel laureate Indian economist Amartya Sen, severely criticized the demonetization move calling it a “despotic action” among other things. Former Senior Vice-President and Chief Economist of the World Bank, Kaushik Basu, called it a ‘major mistake’ and said that the ‘damage’ is likely to be much greater than any possible benefits. Pronab Sen, former Chief Statistician and Planning Commission of India member, called it a “hollow move” since it did not really address any of the purported goals of tackling black money or fake currency. Prabhat Patnaik, a former professor of economics at the Jawaharlal Nehru University, Delhi called the move ‘witless’ and ‘anti-people’.

(I couldn’t help but notice business folks on ‘pro’ side, erudites/scholars/academics on the ‘anti’ side)

Ground Reality

Putting aside personal inconveniences like long waiting times at cash machines & banks, almost fighting for change with the friendly Chennai autos , it’s hard, for a uninitiated, economically naïve citizen like me to gauge the “real” impact of the move. That’s when it dawned on me. I just need to follow the time-tested, proven strategy in such situations: ask your wife. Being the smarty-pants that she is, she gives me a crash course me LMGTFY. At 3 AM. In the tone that is usually reserved for “do the laundry”.

Taking her advice seriously, as any good husband ought to, I google the heck out of it! Did it help? Well, probably. The overwhelming consensus seemed to be “its bad”. Establishments like Forbes, WSJ and the like, were uniform in their critical and underwhelming appraisal of the of the demonetization circa its 1st month anniversary.

Their argument were, and rightly so, predicated on the avoidable inconveniences faced by the people and the sudden economic downturn. There are others who claim these are short-term “pains” worth enduring, given the long term benefits. But that’s a discussion for another time.

The balance?

While the suffering of the people (including the death of dozens) are irrefutable, what struck me the most was the surprising lack of critical appraisal of the primary motive behind this whole demonetization move. Going back to the PM’s address on the 8th of November, demonetization was an effort to stop counterfeiting of the current banknotes allegedly used for funding terrorism. When you google ‘effect of demonetization in India on terror’, there are surprisingly few, if any, from credible news agencies/sources, when compared to the results for searching ‘effect of demonetization in India on people’. Leaving the fairness (or lack there-of) in this disparity, I took it upon me to have a sneak-peek at it.

The Data

Should one stumble upon the list of terrorist organizations in India, although a reasonably long list, one might spot the most active ones operate in Kashmir and the NE. This discussion is strictly going to be about Kashmir.

I scraped data from multiple Indian news channels, in an attempt to offset their (often clear) political leanings  and affiliations. NDTV, CNN IBN and Times Now were the channels of my choice. In my opinion (sans the interesting details) NDTV and CNN IBN have had clear political leanings.

I scrape their data from (roughly early-mid) August to 28th December 2016 for all results featuring under the tag ‘kashmir’. The data lends itself to the usual text cleaning and processing techniques, details of which won’t be discussed here.

Now we need to define the ‘terror related’ terms. What’s to be considered as ‘related’ is subjective and contentious. To ensure an objective appraisal and remove personal biases, the following list is used (this was the first search result for ‘terrorism related words’ in www.google.nl).

No fancy math is required to count the number of times these ‘words’ occur in the corpus and plotting them.

If you wonder if the data is ‘correct’, we can look at some of its features.

These are the top 15 terms, in terms of frequency of occurrence.

words associated with ‘kill’ correlation of words associated with ‘kill’ most frequent terms
terrorist 0.41 armi
soldier 0.36 attack
burhan 0.34 chief
wani 0.33 day
encount 0.31 forc
hizbul 0.24 govern
mujahideen 0.24 india
forc 0.23 jammu
near 0.23 kashmir
armi 0.19 kill
two 0.19 minist
injur 0.18 modi
one 0.18 pakistan
attack 0.17 said
clash 0.17 say

A visual representation of the word associations from articles in NDTV is shown below.

word_associations

(If the text in the plot is not visible clearly, click on it to get an enlarged version)

This seem to make sense. This the kind of news I have grown up seeing in the news time and time again.

Did demonetization address its primary objective?

We plot the occurrences of these ‘words’ in chronological order.

overall

The plots show the concentration of each of the ‘words’. Higher the concentration, black-er the points and Lower, gray-er. In each of the 3 channels, we see more gray areas after the demonetization (the vertical red line). So there is some ‘correlation’ between the demonetization and number of occurrences of the words.

Its fair to assume that had there been any untoward incidents, the agencies would have reported them. So it is to assume that there has been no ‘sudden’ change in the way of reporting from Kashmir. Even if there were, its unlikely to have affected multiple news agencies at the same time.

The above plot gives the ‘correlation’ between demonetization and the reduction in the number of incidents in the Kashmir area. It has to be stressed that the interpretation of the plot uses the term ‘correlation’. It’d be the personal choice of the reader to (or not) extend it beyond correlation.

Although the plot shows more ‘grayness’ (less ‘blackness’) after the demonetization, its possible to visualize it better. A weekly aggregate of the occurrences of the words would help present the trends better, at the cost of ‘reduced’ generalizablity stemming from reduced data points.

weekly_agg

The blue and the red lines show the weekly aggregates of ‘words’ before and after demonetization respectively. All the 3 channels seem to have a noticeable reduction post demonetization. This, again, shows the correlation between demonetization and number of incidents from the Kashmir region.

(Note for the statistically inclined: Violating the independence and sample size criterion of the famed frequentist hypothesis tests, the data reveal a ‘statistically significant’ difference between the pre and post demonetization weekly aggregated word count)

Just like how some children’s groups say they have anecdotal evidence of demonetization’s (positive) repercussions on child labor in areas such as Mumbai, Delhi and other urban centers, we can adopt a similar stance. (But its not statistically valid!)

It’s interesting to notice that post demonetization, not only has the “words” gone down, even the overall count of news articles have gone down.

quieter

What does this mean?

Based on the data of news articles from in 3 Indian news channels, during the period of early August to late December 2016, there seem to be a ‘short-term’ anecdotal evidence (correlation) between demonetization and its primary purpose of reducing the terror funding (to reduce terrorist activities in India). This is indirectly inferred from the reduced the number of untoward ‘incidents’ in Kashmir. This not a statistically proven result yet, partially due to the recency of the event (read limited data post demonetization). Its worth revisiting this analysis a year from now, gather enough data and use stats to arrive at a statistically sound conclusion. The long term, seasonal and random variations of this ‘time series’ would to be accounted for, in such a long term study. While we’d have to wait a few months to gather data and crunch those numbers, the next best we could do look ‘back’ to assess the long term and seasonal behaviour in the past, and examine if the post demonetization data points are any different.

Long term

Here we define ‘long term’ as a long enough term to see the trends and seasonality. Of the 3 news channels investigated earlier, only one, CNN IBN provides access to their long term data. We gather data dating back to summer of 2012. This gives us 4 ‘full’ years of data, enough to, atleast visually, spot trends. (Given the lack of an alternate news channel’s long term data, this data is likely to retain CNN IBN’s political leaning)

So how does it look like?

cnn_ibn_longterm

Its not entirely clear, but one might notice ‘dips’ , especially towards the end of each year. Does this mean the baddies take a X-mas New Year vacation? Or is it the journos? A better way to visualize this data would be the aggregated plots, and this time, we aggregate them monthly.

cnn_ibn_longterm_aggs

Indeed we see clear dips towards the end of the year. It seem to gradually increase, till it hits max in the summer. While the flippant sounding vacation theory might warrant circumstantial merit, I’d personally think after hard work in the summer months, the baddies stay indoors during the winter months- there is no proof for that though. But my point stands.

cnn_ibn_longterm_aggs_dated

It so happens that while the baddies began enjoying their warm cuppas in make-shift winter homes in disputed territory, the demonetization happened. So its almost impossible to link demonetization to the immediate reduction of terror in Kashmir. But to understand the mechanics of this, lets decompose this data into its components.

cnn_ibn_longterm_yearly

This is the year-on-year stack up. The below is the month-on-month stack up.

cnn_ibn_longterm_monthly

A real key to understanding the mechanics is to break it down to its components, aka trend, seasonality and random components. And this is how it looks.

cnn_ibn_longterm_components

One may observe the long term increase in the terror related words in Kashmir related news. Reasons for this are aplenty, but that’s a discussion for another time. The seasonality reveal spikes in the summer and dips in the winter, reinforcing the ‘baddies chilling it out during winter’ theory.

Is this December any different due to demonetization?

This is the million dollar question, isn’t it? As discussed earlier, it would be difficult to attribute demonetization, but nevertheless, we can hold out December’s values, forecast for its value based on historic data and compare the two. (This comparison doesn’t provide statistical evidence for/against the effectiveness of demonetization on Kashmir terror)

forecasts_and_intervals

Standard models were used to forecast the values for December. All the models have their point forecast in the 350-380, the actual value of December is shown by the red dotted line (value = 118). This is well below the forecasts of all the models. But as the plot shows, the actual value is within the (not the 80%, but the) 95% confidence interval, meaning that the actual value is close to the lower echelons of the least expected values.

Interpretation: The odds of the actual December values being so low, is quite low.

What caused it to be low: Hard to isolate, as the confounding variables are aplenty. I encourage the reader to draw their own conclusion.

Political affiliations of this channel?

Does the data suggest political leaning/affiliation (if any) of CNN IBN? If one assumes the news coverage, naming and content are to reflect the channel’s political ideology, then the answer is… probably!

Take a look at the number of news articles regarding Kashmir from summer 2012. While it has a trend, it seem to take a different trajectory, come mid 2014. What happened mid-2014 you ask. Well, we had a new PM swearing in on 26th May 2014.

cnn_ibn_longterm_politicalaffiliation

The red line marks the date of Modi becoming the PM. And immediately after, the trajectory changes. Its fair to ask if there were real terror attacks in Kashmir mid-2014.

This is what the list of terrorist incidents in India wiki page suggests.

wiki

No incidents post April 2014. What’s more surprising is that when there was an attack in Kashmir on the 28th of April injuring 18, the number of news articles flowing out of CNN IBN is significantly lower than an incident free month! It does raise the question of Wikipedia’s credibility as a source of news.

Turning to another source managed by the Institute for Conflict Management, the South Asian Terror Portal, we see that its not the best designed page, astronomically challenging to get any meaningful data out of it, and specifically designed to shoo away all the page visitors… all hallmarks of an underfunded agency. We consult the Jammu and Kashmir Data Sheets and explore  major incidents of terrorist violence in Jammu and Kashmir: 2014, suicide attacks in Jammu and Kashmir, 2014 and explosions in Jammu and Kashmir, 2014.

All of those suggest:

  1. No major or reported attacks in May 2014.
  2. Number of Indians killed and injured in 2014 is less than that of 2013 (yet news channel’s reporting seem to suggest otherwise)
  3. Similar number of suicide attacks in 2013 and 2014 (yet news channel’s reporting pattern suggest higher ‘activity’ 2014)

Its also interesting to note that the channel’s reporting levels about Kashmir has remained ‘noticeably’ higher post may-2014, although the number of fatalities have been on the decline.

I’ll encourage the reader to draw their own conclusion from the above discussion.

Summary

So what did we find out:

  • Baddies feel cold in the winter, just like we do.
  • And when they feel cold, they don’t feel like getting out of the coziness of the makeshift bunkers.
  • With no new notes to buy supplies, they are probably sharing meals, just like we do. But they still cannot share meals in HSB because…
    hsb
  • Demonetization has coincided with the annual winter vacation of the bad guys.
  • A more-than-expected decline in the number of ‘terror words’ in news reported by CNN IBN in December 2016.
  • Hard to relate demonetization to this decline, as the decline could well have been due to baddies slacking in their work, or other random stuff.
  • When it comes to being fair and neutral in news reporting, data suggests that CNN IBN is the new age Steve Bucknor
  • SO’s steaming hot idli and sambar combo is more tempting than adding few more points of this conclusion

4 Comments

  1. It’s an interesting attempt. In terms of media coverage, though, I’d say it’s about news cycles too. The dip in May 2014 could be because Modi’s election was the big news story. In Nov 2016, demonetisation was the bigger story and therefore, fewer instances of indicator words on the sites. What you are absolutely right about is the bias CNN-IBN has and perhaps wears on its sleeve. It may be useful to look at social media mentions from that area about indicator words since this data may be coming from people directly affected without the filter of media.
    Great read, though!

    • Srisai Sivakumar

      I initially thought of doing it the social media way. but the problem is that twitter allows access only to a few days of tweets, not longer. So had to find an alternate way of getting the data

Leave a Reply