Sunday, November 18, 2018

17/11/2018: California Rooftop Solar Mandate: An example of bad groupthink?


In recent news, California legislators have done a gimmick-trick that has earned the state loud applause from the environmentally-minded consumers and activists: California Energy Commission (CEC) recently voted 5-0 to add a new provision to the state’s building code. This includes a requirement that from 2020, all new house and multi-family residences construction of three stories or fewer, along with all major renovations, must be built with rooftop solar panels. Given that the state currently builds ca 113,000 housing units a year, and rising, this should increase significantly already existent solar generation capacity from 15% of the housing stock, currently.

Solar being mandated on virtually all new houses? Sounds like a renewables nirvana, especially given the fact that the state has huge solar generation potential due to its climate. But, as commonly is the case, there is a catch. Or two... or many more... And this means that California's latest policy mandate may be a poor example to follow, and potentially, a bad policy mistake.

Here are the key reasons.

Rooftop solar is about as effective in reducing emissions as waving a broom into the smog. UC Berkeley’s Severin Borenstein argued this in his note to CEC Commissioner (http://faculty.haas.berkeley.edu/borenste/cecweisenmiller180509.pdf). Note: Borenstein also alleges that CEC has failed to involve experts in energy economics in its decision making process - something that is not a good policy formation practice.

UC Davis economics professor James Bushnell accused CEC of “regulatory groupthink.” (https://energyathaas.wordpress.com/2018/10/22/how-should-we-use-our-roofs/) and offered an alternative to roof solar that can generate far greater environmental benefits. There are, of course, other, more efficient ways for deriding emissions, including: mandating more urban density, raising home and cars efficiency standards, expanding the renewable energy mandate, improving grid efficiencies and transmission expansion, and so on. Once again, CEC did not allow for any independent assessment of the proposed plans economic and environmental impacts.

There is an opportunity cost involved in roof solar: California has a state-wide mandate to achieve 50% renewables generation by 2030. Putting more if this target onto roof solar is simply moving generation capacity from one source to the other. Because too top solar is roughly 4-6 times more expensive than industrially-produced renewables, the substitution involves a dramatic reduction in economics of scale. This will raise the overall cost to California of reaching its 2030 target.

Another opportunity cost, this time much more tangible and immediate than 2030 targets is the problem of California grid ability to swallow all the solar generation being put into place. California has to routinely dump excess solar energy supplies during peak generation times, because it is failing to find buyers outside the state. Worse, given the scale of each roof top generation unit, solar electricity from the roof tops cannot be controlled by the grid companies, because smart inverters needed to do this are too expensive for small scale generators.

There is an argument, however, that economics of scale will kick in from a different side: mandating such a huge increase in atomistic (house-level) installations can result in more innovation and lower costs of new technologies going forward. This means that while costs might be high up front, they can potentially be deflated faster over time than absent the mandate. The same argument might hold for improvements in storage.

Worse yet, solar from one roof panel household competes with solar from another roof panel household. All roof top panels generating at virtually the same time across the same time zone state will be simply bidding down the cost of solar during peak generation, not peak demand. Here is an exchange from two experts on this:



Last, but not least, California roof top solar requirements will add new cost, to the housing in a state that is already in the middle of an atrocious housing crisis. CEC own analysis, not tested by any peer review, implies that homeowners are likely to face additional costs of ca $8,000-12,000. Over the depreciation cycle for housing stock, this is likely to translate into $15,600-$23,400 in current dollars (inflation-adjusted, using 2% inflation rate) increase in the cost of housing per household, once property taxes on new build values are factored in. With average house price in California in excess of $420,000, this is equivalent to raising house prices 3.75-5.57 percent. Of course, CEC promises savings that, according the Commission analysis will be net of higher costs. Problem is, no one actually tested these claims, and we simply do not know how the costs of switching all this roof top solar into the grid are going to be distributed across the households.

Macro level view:

Then there is macro level analysis of the solar energy benefits and costs. And California does not come out pretty in this.

A new NBER paper, tiled "Heterogeneous Environmental and Grid Benefits from Rooftop Solar and the Costs of Inefficient Siting Decisions" by Steven E. Sexton, A. Justin Kirkpatrick, Robert Harris, Nicholas Z. Muller (NBER WP 25241, Nov. 2018: https://www.nber.org/papers/w25241.pdf) looked at "federal and state policies in the U.S." These policies "subsidize electricity generation from 1.4 million rooftop solar arrays because of pollution avoidance benefits and grid congestion relief. Yet because these benefits vary across the U.S. according to solar irradiance, technologies of electricity generators, and grid characteristics, the value of these benefits, and, consequently, the optimal subsidy, are largely unknown."

What does this mean? Across the U.S., "policy, therefore, is unlikely to have induced efficient solar investments." The authors provide "the first systematic, theoretically consistent, and empirically valid estimates of pollution damages avoidable by solar capacity in each U.S. zip code". The also link "these external benefits to subsidy levels in each U.S. state, and [estimate] the share of these benefits that spillover to other states." Finally, the authors measure "the energy value of capacity across the U.S. and the value of transmission congestion relief in California."

So what do they find? "Environmental benefits are shown to vary considerably across the U.S., and to largely spillover to neighboring states." Which is not a bad thing in itself, but it also means that some states pay for benefits accruing to other states. These transfers are not voluntary to the payers for solar - the households.

Furthermore, "subsidy levels are essentially uncorrelated with environmental benefits contributing to installed capacity that sacrifices approximately $1 billion per year in environmental benefits." Which, broadly-speaking means that subsidies for rooftop solar are not a great way to achieve environmental benefits.

"...California rooftop solar is shown to generate no congestion relief." Or, as noted above, there are severe grid-related costs involved in rooftop solar in California, the state that decided to mandate it.



Putting more detail on the NBER paper: "Total benefits of solar generation—inclusive of energy values — are estimated to be greatest in the Midwest and Mid-Atlantic. They are least in the West, and particularly the West Coast, where approximately two-thirds of systems are located." Why, given the fact that sunshine is more abundant in California than in the MidWest or Mid-Atlantic?


"These differences are primarily attributable to heterogeneity in marginal responding fossil generation." Oh, wait, that is right: the more solar you put in, the more back up generation you need. And that is before you account for the solar installation possible effects of increasing demand for electricity as the second order effect.

"In California, we find no evidence that rooftop solar capacity systematically relieves congestion. Approximately two-thirds of the 900,000 rooftop solar arrays is located upstream from transmission bottlenecks, contributing to congestion rather than relieving it. If capacity were efficiently allocated, congestion relief benefits in California would have been no more than $15 million in 2017—approximately 7% of total energy value."

Cycle back to that California rooftop solar mandate. Does it really make any environmental sense? Because economics-wise, it does not appear to offer much more than a hype and a pump scheme.

Saturday, November 17, 2018

17/11/18: Nine in Ten in the Red: Asset Markets YTD Returns Signal Risk Repricing


According to a recent research note from the Deutsche Bank, 89% of global macro assets are posting losses on year-to-date basis. This is the highest level of losses in more than a century.


Given the scale of financial risk mis-pricing in equities and bonds markets in the post-QE period, we are likely to witness more downward movement in the assets valuations in months to come. A gradual deleveraging that the market trends have been supporting so far remains highly incomplete and requires more pronounced re-pricing of assets to the downside.

Read more on this here: http://trueeconomics.blogspot.com/2018/11/161118-horsemen-of-financial-markets.html

Friday, November 16, 2018

16/11/18: Student Debt Hits Another High in 3Q 2018


Bloomberg @business just now posted that the student loans debt in the U.S. has increased USD37 billion to USD1.44 trillion at the end of 3Q 2018:


And, Flows of student debt into serious delinquency - 90 or more days - rose to 9.1% from 8.6% in 2Q.

This is somewhat at odds with the Fred database which shows Student Loans debt at USD1.5636 trillion in 3Q 2018, up ca USD33.23 billion on 2Q 2018:


While the NY Fed report is already alarming in both delinquencies rates dynamics and overall debt dynamics, the FRED data that includes securitized debt volumes is even more worrying.

By its very nature, student loans debt impacts the segment of the population (younger workers) who are in the need to fund their housing needs just as their careers are only starting (with associated lower earnings). These younger households also need financial resources to achieve sufficient mobility to better match jobs offers and career prospects to their abilities and needs. Student loans fall heavily onto the shoulders of younger families with growing housing needs, healthcare demand and funding calls from childcare. In other words, student loans debt is potentially crippling those households that are demographically going through the period when enhanced mobility and financial resilience are necessary to secure better life-cycle employment and family outcomes.

16/11/18: The Horsemen of the Financial Markets Apocalypse


My column for the Manning Financial  is now available here: http://issuu.com/publicationire/docs/manning_financial_november_2018?e=16572344/65613030.



Thursday, November 15, 2018

15/11/18: BIS on payments systems and cryptos / blockchain


On November 1, Agustín Carstens, General Manager, Bank for International Settlements delivered a pretty punchy speech on the topic of payments systems evolution in modern age of digital technologies. Punchy, in the sense that much of it is focused on, indirectly, enlisting the evidence as to the lack of the markets for the blockchain and cryptocurrencies deployment in the payments systems at the wholesale and retail levels.

Take the following:  "One of the most significant developments in the evolution of money has been its electronification and, more recently, digitalisation. ...Realtime gross settlement (RTGS) systems for interbank payments, ...emerged in the 1980s. ...RTGS systems allow banks and other financial institutions to send money to each other with immediate and final settlement. They are typically operated by central banks and process critical (read: high-value) payments to allow for the smooth functioning of the economy. Today, the top interbank payment systems in the G20 countries settle more than $17.5 trillion a day, which is over 50 times a working day’s global GDP. ...Given the technology cycle, many central banks are currently looking at next-generation RTGS systems to offer more robust operations and enhanced services."

What does this imply for the world of cryptos? In simple terms, there is no market for cryptos as platforms for interbank payments settlements - the market is already served and the speed of services, cost and security are underpinned by the Central Banks.

Next up: retail payments systems.

Starting with back office: "For retail payment systems, ...in Mexico consumer payments operate at the same speed as interbank payments... The beneficiary of a payment is credited money in near real time. That is, if I were to send you money from my Mexican bank account, you would see the funds in your Mexican bank account in 15 seconds or less. ...Based on a BIS analysis, fast payment systems are likely to become the dominant retail payment system by 2023."

Again, what's the market for blockchain systems to be deployed here? I am not convinced there is one, especially as payments latency and costs are, to-date, more prohibitive under blockchain systems than using traditional payments platforms.

Front office: Carstens notes the progress achieved in delivering what he describes as "payments ... made using bank account aliases" in Argentina that are instant in time, and the ongoing trend toward development of the front-end payments interfaces, based on "cashless systems – no cashiers, no lines, no cash, no physical payment devices. Amazon and others envision a future where you walk into a store, take what you want, and are automatically billed for the items using facial recognition and artificial intelligence. Though this approach may seem a bit scary, it is less so than having microchips implanted inside us, which some firms are also piloting! To be frank, though, neither of these options – facial recognition or microchip implants – are particularly appealing to me."

Carstens presents the evidence that shows current Advanced Economies already carrying more than 90 percent of wholesale payments via cheap, lightning fast and highly secure centralized RTGS systems, with 75 percent of payments via the same occurring in the Emerging Markets:


Given this rate of adoption, coupled with the evolving technology curve (that enables similar systems to be deployed in smaller settlements), one has to question the extent to which cryptocurrency solutions can be deployed in the payments systems.

Beyond the not-too-optimistic view of the market niche size, cryptos and blockchain are also facing some serious pressure points from already ongoing innovation in centralized clearance systems. "Although much attention has been focused on cryptocurrencies as the “it” innovation in payments, there’s much unheralded innovation going on" in the Central Banks and elsewhere (read: legacy providers of payments). "Central banks have been pushing the boundaries of what technology can achieve for operational robustness, including switching seamlessly between data centres at short notice and synchronising geographically dispersed data centres."

Carstens notes the potential for the distributed Ledger Tech (aka, blockchain based on private, enterprise-level blockchain) in this space, where innovation is also a domain of the centralized players, as opposed to decentralised crypto markets. "One interesting development in the central banking community is ongoing experimentation with distributed ledger technology (DLT) as a means to enhance operational robustness. People often use DLT and Bitcoin interchangeably, but they are not the same! ...DLT is simply a set of processes and technologies that enable multiple computers to maintain collectively a common database. DLT does not mean mining of coins, public ledgers and open networks. And no central bank that I’m aware of is contemplating these properties in its DLT experimentation."

There are some problems, however, for DLT enthusiasts:
1) "...a Bank of Canada study noting that a DLT-based payment system meeting central bank requirements would be similar to what we have today (ie private ledgers, closed networks and a central operator). The difference is that a network of computers would be used to settle a transaction instead of one computer." In other words, there is a case, yet to be proven, that DLT offers anything new to the payments systems to begin with.
2) "The second is an ECB and Bank of Japan study concluding that processing times would be three times longer using DLT versus current systems." In other words, DLT/blockchain cannot deliver, so far, on its main premise: higher processing efficiency than legacy systems.




Carstens sums it up: "My take is that current versions of DLT are not any better than what we already have today."

In other words: DLT/blockchain solutions appear to be:

  • Not necessary: the technology is attempting to solve the problems that do not exist in the payments systems;
  • Inefficient with respect to its core tenants/promises: the technology is inferior to existent solutions and the pipeline of ongoing improvements to the legacy systems.
Which begs two questions that the DLT/blockchain community needs to answer: What niche can blockchain occupy in payments systems going forward? and Is there a sustainable market within that niche that cannot be captured by alternative technologies?

But there is more. Carstens explains: "Cryptocurrencies, such as Bitcoin, Ether and Tether, do not serve the core functions of money. No cryptocurrency is a true unit of account or a payment instrument, and we have seen this year that they are a poor store of value. This then raises the question: what are they?" The answer should be a wake up call for anyone still long cryptos: "From my perspective, cryptocurrencies are, at best, an asset of some sort. Perhaps an asset comparable to a piece of art for those who appreciate cryptography. Buyers of cryptocurrencies are buying into nothing more than a software algorithm. Some firms are trying to back cryptocurrencies with an underlying asset, such as cash or securities. That sounds nice, but it’s the equivalent of making art from banknotes or stock certificates. The buyer is still buying an idea or a concept or, if you will, an asset that is the equivalent of art hanging on your wall. If people want the underlying asset, they might be better served just buying that."

Carstens previously (February 2018) claimed that the #cryptos are “combination of a bubble, a Ponzi scheme and an environmental disaster.”

Nice perspective. If you are an observer. For a holder of cryptos, this is a serious risk. Playing cards in a casino is fun, but it is not investing. Playing investing in the cryptos world is probably the same.


Note: for an even more 'in your face' assessment of the #Bitcoin and #Cryptos, there is ECB's Executive Board member, Benoit Coeure, who called #BTC the “evil spawn of the [2008] financial crisis, per Bloomberg report of November 15 (https://www.bloomberg.com/news/articles/2018-11-15/cryptocurrencies-are-evil-spawn-of-the-crisis-for-ecb-s-coeure).

The reality of #cryptos investments is that they are, empirically, a massively overvalued bet on the largely undeveloped and unproven (in real world applications) technologies that have only tangential relation to the coins currently traded in the markets. It is, in a way, a derivative bet on a future contract.

15/11/18: The 'New Normal' is a Road to another Tech Sector Bust


The VC land of wonders and waste is awash with cash, thanks to a decade-long loose liquidity pumping across the markets by the Central Banks. Just as in the prior iterations of the same (the Dot.Com Bubble and the pre-GFC assets binge), the outrun will be the same as it was before: a crash.

TechCrunch reports that (https://techcrunch.com/2018/11/11/age-of-the-unicorn/):

  • Over the last 5 years, the number of 'unicorns' - startups with valuations in excess of USD1 billion - has grown from 39 to 376 - almost a ten-fold increase
  • The rate of 'unicorns' emergence is accelerating: in 11 months through November 1, 2018, we've added 81 new 'unicorns' to the roster, which means there is now a new 'unicorn' company launched every four days
  • Mega-deals for start ups - funding rounds in excess of USD100 million - are also on the rise, with their frequency up ten-fold on five years ago. "Back in 2013, there were only about four mega rounds a month, but now there are forty mega rounds a month based..." Thus, "starting from 2015, public market IPO has for the first time no longer been the major funding source for unicorn size companies."

As the chart above shows, there has been a power-law acceleration in the trend since mid-2017 and it is now clearly topping the asymptote.

Two countries dominate the 'unicorns' league: China (with 149 count) and the U.S. (with 146 count). Which implies two things: 
  • Given the close links between the PBOC policies, Chinese Government investment strategies and supports, and China's counts of 'unicorns', majority of these start ups are heavily dependent on debt, and political good will. They are sitting ducks for ESG risks and are extremely exposed to political and policy uncertainty.
  • The U.S. 'unicorns' are completely dependent on the markets ability to cycle cash from corporate and financial sectors debt and private equity into start ups funding, and M&As. There is zero rational valuation happening in this sub-sector.
A dramatic shift in risks from tangible tangible technologies (including strongly patentable innovation or defensible market shares) of the likes of Apple and Google toward less tangible, highly price and income elastic SaaS types of product offers is reflecting the massive buildup in valuations risks. This too is reflected in the article, albeit the authors fail to spot the implications. TechCrunch conclusion is perhaps even more alarming that the stats they present. "Mega rounds are the new normal; staying private longer is the new normal; and the global composition of the unicorn club is the new normal." We've heard exactly the same arguments at the tail end of the Dot.Com boom about the absurdly over-valued early internet age companies. We've heard exactly the same arguments about the real estate sector prior to 2008. We've heard exactly the same arguments about tulip bulbs in Amsterdam some centuries ago too. 

'The new normal' is the old road to a bust.

Tuesday, October 30, 2018

29/10/18: Corporate Credit and the Debt Powder Keg


As it says on the tin: despite growth in earnings, the numbers of U.S. companies that are struggling with interest payments on the gargantuan mountain of corporate debt they carry remains high. The chart does not show those companies with EBIT/interest cover ratio below 1 that are at risk (e.g. with the ratio closer to 0.9) for the short term impact of rising interest rates. That said, the overall percent of firms classified as risky is at the third highest since the peak of the GFC. And that is some doing, given a decade of extremely low cost of debt financing.

Talking of a powder keg getting primed and fused…


Saturday, October 27, 2018

26/10/18: De-democratization of our values?


Do you, my friends love that smell of napalm in the mornings? Does it fill you with confidence about the future - your own and that of the rest of the world? For if you do answer 'Yes' to both of these questions, congratulations: you've made it into the American values (on average).

Here is the data from Pew Research on public trust in institutions:
Source: http://www.pewresearch.org/fact-tank/2016/10/18/most-americans-trust-the-military-and-scientists-to-act-in-the-publics-interest/ft_16-10-18_trustinstitutions_overview/.

Democratic cornerstones of political leadership and the media are trusted by less than 50 percent of the Americans, 27 percent to 41 percent. Command & control institution of the military is trusted by 79 percent. Scepticism over coercive power and centralized counter-democratic system based on rank-command trumps all the softy stuff. even educators are more trusted in the settings involving more deterministic decision-making (e.g. medical sciences and sciences) than in more polemical setting (e.g. general education).

But if you thought European are all softy-dofty democratically minded, think again. In fact, there is not much of difference between the American population and the European ones on the same topic:
Source: http://www.pewresearch.org/fact-tank/2018/09/04/trust-in-the-military-exceeds-trust-in-other-institutions-in-western-europe-and-u-s/.

Yes, yes, I hear you, the choir of Patriots: but in a democratic society, military is subject to democratic checks and balances, so why is this degree of confidence in military a matter for concern?

Let me explain why:

  • Firstly, irrespective of the democratic or other checks and balances, military is not a pluralist institution that encourages debates, inquiries, and discoveries through dissent. In other words, while military may be framed into the broader democratic constraints, it cannot by itself be a genuinely democratic institution. There is no democracy in command systems. 
  • Secondly, to anyone pontificating on democratic checks and balances, may I suggest revisiting the entire modern history of the U.S. to identify exactly at what point in time did democratic checks and balances imposed onto military do their jobs before the damages were incurred? In Vietnam? Nope. In multiple military engagements in Latin America? Nope, again. In the Balkans? Not really. In Afghanistan or Iraq? Not at all. 
Thus, in real terms, checks and balances do not define a democratic set of values that put strong public preference in favour of the military ahead of much weaker preferences in favour of the immediate democratic and pluralist institutions, such as politics, media, business, education.


What is happening, thus, as revealed by the above data, is the strong drift in public preferences away from democratic and liberal foundations of our modern states and toward more command and control, more coercive power-based institutions, such as military. To paraphrase one semi-failed leader of the past: it's public values de-democratization, stupid.

26/10/18: Visualizing Mental Health Around the World and Across Time


Mental health is one of the key parameters of the quality of social and personal environment we can think of, and changes in the prevalence and the impact of mental health are hugely important measures of socio-economic evolution. Here is a fascinating set of data visualizations and analytical notes on the state of mental health around the world and across the recent decades: https://ourworldindata.org/mental-health.

There are thousands of interesting, and often non-intuitive, observations one can draw from these interactive charts and from the published analysis.


Friday, October 19, 2018

19/10/18: IMF's Woeful Record in Forecasting: Denying Secular Stagnation Hypothesis


A recent MarketWatch post by Ashoka Mody, @AshokaMody, detailing the absurdities of the IMF growth forecasts is a great read (see https://www.marketwatch.com/story/the-imf-is-still-too-optimistic-about-global-growth-and-thats-bad-news-for-investors-2018-10-15?mod=mw_share_twitter).  Mody's explanation for the IMF forecasters' failures is also spot on, linking these errors to the Fund's staunch desire not to see the declining productivity growth rates (aka, supply side secular stagnation).

So, to add to Mody's analysis, here are two charts showing the IMF's persistent forecasting errors over the last four years (first chart), set against the trend and the cumulative over-estimate of global economic activity by the Fund since mid-2008 (second chart):




While the first chart simply plots IMF forecasting errors, the second chart paints the picture fully consistent with Mody's analysis: the IMF forecasts have missed global economic activity by a whooping cumulative USD10 trillion or full 1/8th of the size of the global economy, between 2008 and 2018. These errors did not occur because of the Global Financial Crisis and the high degree of uncertainty associated with it. Firstly, the forecasting errors relating to the GFC have occurred during the period when the crisis extent was becoming more visible. Secondly, post GFC, the hit rates of IMF forecasts have deteriorated even more than during the GFC. As Mody correctly points out, Fund's forecasts got progressively more and more detached from reality.

At this stage, looking at April and October 2018 forecasts from the Fund's WEO updates implies virtually zero credibility in the core IMF's thesis of a 'soft landing' for the global economy over 2019-2021 time horizon.

19/10/18: There's a Bubble for Everything


Pimco's monthly update for October 2018 published earlier this week contains a handy table, showing the markets changes in key asset classes since September 2008, mapping the recovery since the depths of the Global Financial Crisis.

The table is a revealing one:


As Pimco put it: "The combined balance sheets of the Federal Reserve, European Central Bank, Bank of Japan, and People’s Bank of China expanded from $7 trillion to nearly $20 trillion over the subsequent decade. This liquidity injection, at least in part, underpinned a 10-year rally in equities and interest rates: The S&P 500 index rose 210%, while international equities increased 70%. Meanwhile, developed market yields and credit spreads fell to multidecade, and in some cases, all-time lows."

The table points to several interesting observations about the asset markets:

  1. Increases in valuations of corporate junk bonds have been leading all asset classes during the post-GFC recovery. This is consistent with the aggregate markets complacency view characterized by extreme risk and yield chasing over recent years. This, by far, is the most mispriced asset class amongst the major asset classes and is the likeliest candidate for the next global crisis.
  2. Government bonds, especially in the Euro area follow high yield corporate debt in terms of risk mis-pricing. This observation implies that the Euro area recovery (as anaemic as it has been) is more directly tied to the Central Banks QE policies than the recovery in the U.S. It also implies that the Euro area recovery is more susceptible to the Central Banks' efforts to unwind their excessively large asset holdings.
  3. U.S. equities have seen a massive valuations bubble developing in the years post-GFC that is unsupported by the real economy in the U.S. and worldwide. Even assuming the developed markets ex-U.S. are underpriced, the U.S. equities cumulative rise of 210 percent since September 2008 looks primed for a 20-25 percent correction. 
All of which suggests that the financial bubbles are (a) wide-spread and (b) massive in magnitude, while (c) being caused by the historically unprecedented and over-extended monetary easing. The next crisis is likely to be more painful and more pronounced than the previous one.


Tuesday, October 16, 2018

16/10/18: Data analytics. It really is messier than you thought


An interesting study (H/T to @stephenkinsella) highlights the problems with empirical determinism that is the basis for our (human) evolving trust in 'Big Data' and 'analytics': the lack of determinism in statistics when it comes to social / business / finance etc data.

Here is the problem: researchers put together 29 independent teams, with 61 analysts. They gave these teams the same data set on football referees decisions to give red cards to players. They asked the teams to evaluate the same hypothesis: are football "referees are more likely to give red cards to dark-skin-toned players than to light-skin-toned players"?

Due to a variation of analytic models used, the estimated models produced a range of answers, from the effect of skin color of the player on red card issuance being 0.89 at the lower end or the range to 2.93 at the higher end. Median effect was 1.31. Per authors, "twenty teams (69%) found a statistically significant positive effect [meaning that they found the skin color having an effect on referees decisions], and 9 teams (31%) did not observe a significant relationship" [meaning, no effect of the players' skin color was found].

To eliminate the possibility that analysts’ prior beliefs could have influenced their findings, the researchers controlled for such beliefs. In the end, prior beliefs did not explain these differences in findings. Worse, "peer ratings of the quality of the analyses also did not account for the variability." Put differently, the vast difference in the results cannot be explained by quality of analysis or priors.

The authors conclude that even absent biases and personal prejudices of the researchers, "significant variation in the results of analyses of complex data may be difficult to avoid... Crowdsourcing data analysis, a strategy in which numerous research teams are recruited to simultaneously investigate the same research question, makes transparent how defensible, yet subjective, analytic choices influence research results."

Good luck putting much trust into social data analytics.

Full paper is available here: http://journals.sagepub.com/doi/pdf/10.1177/2515245917747646.