13 Nov 2016

Did the Currency Ban leak lead to spike in Bank Deposits?

The recent currency ban by the Govt has attracted feedback both positive and negative. As expected, the opposition largely slammed the move either from the very beginning (Mamta Banerjee tweeted all in caps, terming this a draconian move) or after waiting for a couple of days (Congress lent cautious support in the beginning, and then changed tone looking to capitalise on inconvenience being faced by common citizens)

These don't matter. Traditional parties (including BJP) oppose almost every Govt decision while in opposition, even if they take similar decision when in power (Eg. GST)

However, when AAP alleges a huge scam, I generally dig deeper since they have occasionally brought out genuine issues based on good data analysis. Kejriwal, in a recent press conference, cited a supposedly huge spike in deposits prior to the Currency-ban announcement by the Govt, and accused Govt of leaking the news to its "friends", who then deposited their black cash stash before the ban (causing the spike), and hence somehow converted their back to white. On social media, a graph (reproduced below) was also shared by a user quantifying this allegation, showing spike in Sep-16 deposits compared to Aug-16, whereas no such spike was seen in 2015. Slam dunk case for the assertion on deposit spike, it would seem. 
Source: Gaurav Shukla

This post is to discredit this assertion.

But let me first let me dispute the inference that such Deposit spike (assuming the assertion is correct) prior to the PM announcement has helped the deposit-holder launder the money. Incorrect. If someone deposits money in the bank, it has to be white, irrespective whether it was deposited before or after the announcement. If black money is deposited in the bank, again irrespective of whether the deposit was done before or after the announcement, IT department can easily come after you. All deposits >50k are tagged by Pan Card, and in any case almost all banking accounts are KYC-compliant. And even if someone laundered black money before the announcement by aggregating multiple individuals (thereby each single deposit will be of small value which IT dept shall evade), he can do that even now, there is a 50 day time window to deposit old notes through the time-tested money-laundering techniques.

So even if "Friends of BJP" were warned beforehand, and they deposited their black cash stash in the banks before PM announcement on Currency Ban, they would be caught. And if it is white money cash that was deposited, it could very easily be deposited after the announcement. And if it is black money deposited by aggregating thousands of individuals, even that is possible after the announcement given the 50-day time window. 

So now let me come back to data on Deposits showing a spike before the PM announcement. Whoever made this graph consciously or unconsciously left out October data - RBI publishes data in multiple places and not everything is synced. Below table presents the same. I have also shown Demand Deposits separately, because that is what I believe leads to the overall spike (as per hypothesis I shall pose later)

Note: My numbers are slightly different from the graph above. I have excluded inter-bank deposits which is the right way to look at the data. But the growth rates are similar

As evident from table above, there is a de-growth of 2.2% in deposits in Oct-16 compared to Sep-16. In fact, demand deposits de-grew by 11% in Oct-16! If indeed "Friends of BJP" were forewarned and they scrambled to deposit their black cash (which anyways wont have helped them escape IT dept as pointed out earlier), naturally deposits (and demand deposits) should have continued to increase in Oct-16 as well. So the assertion, when looking at complete data, doesn't stand true 

But still, the spike seen in Sep-16 is interesting, and needs to be explained. For that, first one needs to understand the data construct. This data is sourced from RBI (you can source it yourself by following directions provided at the end of this post), and represents deposits as on the last working Friday of that month, which might or might not be the last working day of that month. So the numbers we see in the graph (as being widely circulated on SM and reproduced above) and the table above are not for month-end. Oct-16 means 28th Oct, Sep-16 means as of 30th Sep and Aug-16 means as of 26th Aug. This is important. Banks are commercial organisations and almost all of them are listed, and their employees have quarter-end targets as is the case for most other listed companies. In case of banks, deposit base "miraculously" increases significantly on the last working day wherein branch managers "request" their corporate / SME customers to deposit as much money as possible in the bank in their current account (and that is why I have included the demand deposits in my analysis above) on the last working day, just to be withdrawn the next working day. I have seen it happen first-hand in multiple banks where my professional experience has allowed me to have an inside peek. In most cases, these last-day deposits are nothing but money drawn from a working capital account that the corporate / sme customer has with the same/different bank, which also leads to a spike in bank credit (more on that later).

The Sep-16 data in the above table is as of 30th Sep 2016 (the last working Friday of that month), which also happens to be the last working day of that quarter. Hence, I am not surprised by the 5% spike from previous month (and 15% in Demand Deposits), the entire bank machinery works overtime to achieve the quarter-end targets. To validate this hypothesis, I looked at monthly deposits data for last ~8 years (Jan-09 onward) and calculated avg growth rate on a month-on-month basis (94 instances, pretty decent sample size). I found 8 such instances when the last working Friday also happened to be the last working day of that quarter, the below table shows monthly deposit growth rate (and demand deposit growth rate) for these 8 instances.

So, the last time it happened, it was 29th march 2013 when the last working Friday was also the last working day of that quarter, and deposits spiked by 5.6% (demand deposits by 21.7%). On an average, deposits grew by 3.9% over these 8 instances (demand deposits by 15.2%), so a deposits growth of 4.9% (and demand deposits growth of 15.5%) for Sep-16 is nowhere out of the ordinary. To be fair, even Saturdays used to be working earlier but then it used to be a half-working day, and many corporate/sme customers from which these last minute deposits were sourced were also closed on Saturday, so it would be fair to assume that most of such deposit mobilisation was completed on Friday. I would have loved to have Saturday data here, but we don't have it. 

To complete the analysis, the below table compares these 8 instances (last working friday = last working day) with 23 other quarter-end instances (last working friday <> last working day), as well the 63 remaining month-end instances which were not quarter-end in terms of the monthly deposit (and demand deposit) growth rates. 

As the table above shows, for the 8 instances of quarter-end where last working Friday was also the last working day of that quarter, deposits grew by an average of 3.9% (demand deposits by 15.2%) as compared an average of 1.0% avg deposit growth rate (and 3.7% avg demand deposit growth rate) for the 23 other instances of quarter-end where last working Friday was not the last working day of that quarter.

Another way to validate this hypothesis is to also check the bank credit growth rate, which should also exhibit similarly different growth rates depending on whether the last working Friday is the last working day or not. Turns out, it does! See the table below (I have included the Deposits growth rate from previous table for comparison).

Hence, for the 8 instances of quarter-end where last working Friday was also the last working day of that quarter, bank credit grew by an average of 4.4% as compared an average of 1.9% for the 23 other instances of quarter-end where last working Friday was not the last working day of that quarter. 

I would have loved to compare only the working capital credit growth rate (and not the overall bank credit) with the demand deposit growth rate, since that it where the real spike happens. But we do not have that data.

Finally, I also need to state that bank deposits is not equal to currency in circulation, i.e. the increase in deposits does not mean fresh currency earlier residing as cash has been now deposited in the bank. Total currency in circulation in India is less than Rs 20 tn, whereas the deposits are well above Rs 100 tn. This similar to GDP multiplier effect, which people interested in macro-economics can research on their own. But very briefly (and crudely), if you deposit Rs 100 in the bank, the bank lends that to a business who uses it to generate output which, partially, gets converted to cash in the hands of the business as well as its employees, which in turn will again deposit the same in bank. So original Rs 100 deposit leads to a  much larger deposit in the bank. So the Rs 5 tn increase in bank deposits in Sep-15 doesn't mean Rs 5 tn currency has been taken out of circulation and deposited in the banks. This is not exactly how it happens, but you get the gist. 

Directions to source data used for above analysis: 
1. Goto https://dbie.rbi.org.in/DBIE/dbie.rbi?site=home
2. Click on "Business of Scheduled Banks" on left side
3. Export data in excel by clicking on the relevant button at the top of the table
4. In the excel file downloaded, use "2.1) Aggregate deposits", "2.1.1) Demand deposits" and "7) Bank credit" for the three data series used in the above analysis.    

"In God we trust, rest all must have data"

Edit: There seems to be an explanation for Sep-16 deposit spike that 7th Pay Commission arrears were released on 31st Aug which caused the spike. News search suggests that number to be Rs 346 bn, far less than the spike of ~Rs 5 tn, so that doesn't seem to explain this. Unless the number of Rs 346 bn is wrong. I still think the last working Friday being the last working day of that quarter is the correct underlying reason. Further, added credit for the first graph.