Wednesday, May 24, 2017

23/5/17: Eurocoin: Growth Momentum Slips Marginally in April


A quick update to the old-running series: Eurocoin, the leading economic growth indicator for the euro area, published by CEPR and Banca d'Italia posted another (second in a row) moderation, falling from 0.7 in March 2017 to 0.67 in April. The indicator remains at the upper range of growth for the current upside cycle, and within lower range of growth compared to previous upside cycle:


On the drivers side, stock markets valuations helped to push growth forecast higher, while a slowdown in industrial activity pushed growth expectations lower. In other words, absent the financial assets impact, growth indicator would be much lower.

While euro area overall HICP was at 1.9% in April (bang at the upper range of ECB's target), 12mo trailing average inflation rose to 0.8% from 0.7% in March. Which means the ECB has moved out of the 'policy corner' and is now positioned to start unwinding assets purchasing programs. It will proceed gradually and at a later date, due to political, not monetary reasons.

Meanwhile, although Eurocoin averaged 0.72 in 1Q 2017, actual growth came in at (first estimate) 0.5%. This marks the largest gap between Eurocoin and actual growth since 2Q 2014. This is hardly surprising. In general, the gap between leading indicator-implied growth forecast and actual growth outrun is usually wider during periods of elevated uncertainty about the economy, and especially when financial economy takes over as a major contributor to overall economic growth outlook.

Tuesday, May 23, 2017

22/5/17: Eurogroup and Greece: Wrestling Defeat from the Claws of Victory


Today's Eurogroup meeting on Greece ended in no agreement and extends the current tranche negotiations into June 15, the date of the next Eurogroup meeting.

For the background:

The key sticking point so far is the scheduling of future primary surpluses (budgetary surplus before the debt servicing costs are factored in). The Eurogroup insists on these surpluses running at 3.5% of Greek GDP for the first 5 years following 2018, declining to 2% or 2.2% (depending on the version of the draft agreement) for 2023-2060. 

In very simple terms, such commitments are absolutely bogus (and dangerous). They are bogus because there is absolutely no way anyone can project growth rates out to 2060 from today that can be in any way reasonably accurate to predict primary surpluses. They are dangerous, because they will shackle Greek governments to running buffer funds to compensate for possible recessionary and non-cyclical shocks to the primary surpluses. These buffers will imply underinvestment within the Greek economy (public investment) over the long term. Which, of course, will damage the Greek economy and increase the risk of non-compliance with the deficit rules.

Here is how unrealistic the current proposed targets are. Consider, first, IMF projections (April 2017 data) for primary surpluses over the next 5 years (2018-2022). Remember, Greek target (grey line) is 3.5% for that period:

With exception of Italy, no other advanced euro area economy comes even close to the proposed target. And no one is making a case that Italy running these surpluses is somehow consistent with structurally strong growth expectations over the period.

Now, consider past and present performance, based on 10 years windows. For 10 years window, Greek target surplus is 2.85% per annum:

The view is a bit brighter. 

In the 1990s, two countries managed to run surpluses at or above the target set for Greece forward: Belgium and Ireland. Both countries were recovering from substantial fiscal crises of the late 1980s-early 1990s.  But, unlike Greece today, both countries benefited from exogenous shocks that boosted significantly their surpluses and growth: Belgium gained substantial income transfers from growth of the EU institutions, and Ireland gained from a large scale FDI boom. Neither country needed to run large scale public investment programmes financed from own (internally-generated) funds. 

In the 2000s, Belgium continued to run large surpluses and it was joined in this by Finland. Belgium surpluses drivers remained the same, while Finland carried out substantial fiscal consolidation in the wake of the early 1990s crisis timed perfectly to coincide with rapid economic growth in the economy. 

In simple terms, no advanced euro area economy has managed to run surpluses expected of Greece at the times of adverse economic growth conditions or immediately after a major recession.

As I noted in the earlier post on the Greek economy (see http://trueeconomics.blogspot.com/2017/05/18717-greece-in-recession-again.html), the state of Greek economy has been so highly uncertain over the last few years, that any projections 3-4 years out from today are simply an example of a delirious wish-for-thinking. In this environment, setting targets out to 2060 is absurd, and dangerous, for it commits Greece to targets that may or may not be to the benefit of the Greek economy and sets up the euro area fiscal policy architecture for a failure at the altar of extreme conviction in technocratic targeting. 

Monday, May 22, 2017

22/5/17: The Economist Calling a Bubble in Cryptos? Why not...


A very interesting chart from The Economist from last week providing evidence on rapid evolution of cryptocurrencies:
Source here

In basic terms, the value of cryptocurrencies market has risen to over USD60 billion, tripling within less than 6 months of 2017, while trading across cryptocurrencies markets has increased tenfold to ca USD 2 billion per day (average) and 38 initial coin offers have been launched in 2017 so far.

What is also notable is that Ripple is now on par with Ethereum and their combined valuation is now a challenger to Bitcoin.

Uncharacteristically for The Economist,  the publication that never sees a bubble until it pops is calling a bubble in the cryptos. Perhaps due to a freshly acquired consciousness of empiricism or due to the publication's innate distaste for anything not-state-centric. Still, given the exponential growth so far this year, cryptos are overdue a major correction. When and how will it be triggered is anyone's guess.

22/5/17: U.S. Autoloans Market: Careless Lending, Defaulting Buyers


Auto loans are now coming through as a growing concerns area in terms of U.S. household credit. Auto loans originations have risen, in total volume from $123.9 billion in 1Q 2016 to $132.4 billion in 1Q 2017, an all-time high for 1Q period on record. Total volume of auto loans debt outstanding is at $1,167 billion, up on $1,071 billion in 1Q 2016 and at an all-time record. Year on year growth in auto loans is at 9%.


However, origination has been more subdued in 1Q 2017 for subprime loans, with issuance for credit score below 620 falling to $25.9 billion in 1Q 2017 compared to $26.9 billion a year ago. Likewise, near-sub-prime originations (credit scores 620-659) also declined, from $16.1 billion in 1Q 2016 to $15.6 billion in 1Q 2017.




However, owing to rapid growth in recent years in sub-prime originations, auto loans currently exhibit third highest rate of delinquencies across all forms of household debt, with 3.82 percent of all auto loans currently 90+ days delinquent, the highest since 1Q 2013 and up on 1Q 2016 reading of 3.52 percent.

As noted in a recent Bloomberg article (see https://www.bloomberg.com/news/articles/2017-05-22/subprime-auto-giant-checked-income-on-just-8-of-loans-in-abs), much of the problem arises from sloppy, or outright careless, origination by some key lenders. 

22/5/17: U.S. Public Pensions System: Insolvent to the Core


A truly worrying view of the U.S. public sector pensions deficits has been revealed in a new study by Joshua D. Raugh for Hoover Institution. Titled “Hidden Debt, Hidden Deficits” (see http://www.hoover.org/sites/default/files/research/docs/rauh_debtdeficits_36pp_final_digital_v2revised4-11.pdf) the study opens up with a dire warning we all have been aware of for some years now (emphasis is mine):  “Most state and local governments in the United States offer retirement benefits to their employees in the form of guaranteed pensions. To fund these promises, the governments contribute taxpayer money to public systems. Even under states’ own disclosures and optimistic assumptions about future investment returns, assets in the pension systems will be insufficient to pay for the pensions of current public employees and retirees. Taxpayer resources will eventually have to make up the difference.”

Some details: “most public pension systems across the United States still calculate both their pension costs and liabilities under the assumption that their contributed assets will achieve returns of 7.5–8 percent per year. This practice obscures the true extent of public sector liabilities.” In other words, public pension funds produce outright lies when it comes to the investment returns they promise to generate. This, in turn, generates delayed liabilities that are carried into the future, when realised returns come in at some 3-4 percent per annum, instead of promised 7.5-8 percent.

How big is the hole? “In aggregate, the 564 state and local systems in the United States covered in this study reported $1.191 trillion in unfunded pension liabilities (net pension liabilities) under GASB 67 in FY 2014. This reflects total pension liabilities of $4.798 trillion and total pension assets (or fiduciary net position) of $3.607 trillion.” This accounts for roughly 97% of all public pension funds in the U.S. Taking into the account the pension funds’ penchant for manipulating (in their favor) the discount rates, the unfunded public sector pensions liabilities rise to $4.738 trillion.

“What is in fact going on is that the governments are borrowing from workers and promising to repay that debt when they retire. The accounting standards allow the bulk of this debt to go unreported due to the assumption of high rates of return.”

Actually, what is really going on is that the governments create a binding contract with their employees to loot - at some point in the future - the general taxation funds to cover the shortfalls on these contracts. How much looting is on the pensions liabilities? Take the unfunded liability estimate of $4.738 trillion. And consider that in 2014, total revenues collected by state and local governments stood at $1.487 trillion. Pensions deficits alone amount to 3.2 times the underwriters’ income. In household comparative terms, this is like having a full 100% mortgage on a second home, while still running a full 100% mortgage on primary residence (day-to-day expenses).

Or, put more cogently, the entire system is insolvent. And is getting more insolvent, the longer the local and state governments refuse to use more honest accounting models.

Couple of charts to illustrate




CHART 2: State Contributions: Actual vs Required to Prevent Rise in Unfunded Liability

Now, observe in the above: the distance between the green triangle (required contributions) and the blue dot (actual contributions) is the gap in public pensions funding that has to be extracted to make the contracts whole. This will either have to come from tax hikes or from increased contributions from the public sector workers or from cut in future benefits to these workers. Or from all three.

In a range of the states, e.g. California, New Jersey, Illinois, etc we are already facing draconian levels of taxation, and falling real incomes of private sector workers. In a range of other states, municipal and local taxes are high, while the cost of living increases are swallowing income growth. In other words, there is not a snowball’s chance in hell these gaps can be funded from general taxation in the future.

When all ameliorating assumptions are made (to the upside for public pensions schemes), Raugh concludes that “despite markets that performed well during 2009–2014, state and local government pension systems are still underwater by $3.4 trillion. With relatively poor performance in fiscal years 2015 and the first part of 2016, this figure is likely to be even larger today. Finally, the report reveals the extent to which state and local governments are in fact not running balanced budgets. While they contribute 7.3 percent of their own-generated revenue to pensions, the true annual ex ante, accrual-basis cost of keeping pension liabilities from rising is 17.5 percent of state and local budgets. Even contributions of this magnitude would not begin to pay down the trillions of dollars of unfunded legacy liabilities.”

Yes, the entire system of public pensions is insolvent. No surprise there. And there is not enough fiscal space to recover from that insolvency without cutting benefits, raising taxes and hiking employee contributions. No surprise there either. Finally, although Raugh does not say so himself, it is pretty clear that there is zero will on either side of the Washington’s political divide to do anything tangible to address the problem.


Note: you can read a series of previous posts covering various sides of household debt in the following threads: Total Household Debt http://trueeconomics.blogspot.com/2017/05/19517-us-household-debt-things-are-much.html; U.S. Social Security Insolvency  http://trueeconomics.blogspot.com/2017/05/19517-reminder-social-security-is-only.html, and Student Loans Explosion http://trueeconomics.blogspot.com/2017/05/21517-student-loans-debt-bubble-is.html).

21/5/17: Student Loans Debt: The Bubble is Still Inflating


Having covered the latest news on the U.S. household debt continued explosion (see http://trueeconomics.blogspot.com/2017/05/19517-us-household-debt-things-are-much.html) and the ongoing deepening of the long term insolvency within the U.S. Social Security system (here: http://trueeconomics.blogspot.com/2017/05/19517-reminder-social-security-is-only.html), let’s take a look at the second largest source of household debt (after mortgages): Student Loans.

According to the data from the New York Federal Reserve, 1Q 2017 total volume of student loans outstanding in the U.S. was USD1.344 trillion, up on USD 1.310 trillion in 4Q 2016, marking the highest level of Student Loans debt in history. However, the Fed methodology does not include some of the more predatory types of loans extended to students.  This means that other sources report student debt at the end of 2016 to be between USD 1.44 trillion and USD 1.5 trillion (see for example https://studentloanhero.com/student-loan-debt-statistics/).

Setting aside the issues relating to data reporting, even by the official U.S. Fed standards, student loans debt is almost double the U.S. households’ credit cards debt, and is more than 10 percent higher than combined credit cards and HE revolving debt volumes.

Crucially, default rates on Student Loans are currently higher (at 11%) than for any other form of debt (credit cards defaults, second highest, are at around 7.45%).


With an average debt load of over USD36,000 per student, the expected Fed rates hikes through 2017 alone are likely to take some USD 270.00 per annum from household budgets already under severe strain from low income growth, and sky high and rising rents.

Meanwhile, the U.S. bankruptcy code now excludes Student Loans from protection, courtesy of 2005 Congressional decision, presided over by Joe Biden. Of course, Biden’s political machine was supported by one of largest student loans underwriter, the MBNA. President Obama promised to undo Biden’s changes to the bankruptcy code, but in the end did absolutely nothing to keep his promise and, in fact, made matters worse (https://www.bloomberg.com/news/articles/2015-10-14/obama-administration-hits-back-at-student-debtors-seeking-relief).  Subsequently, during his Presidential election campaign, Donald Trump said “That's probably one of the only things the government shouldn't make money off -- I think it's terrible that one of the only profit centers we have is student loans.”  Since election, however, the Trump Administration is yet to do anything on the issue.

In simple terms, American students have no friends in high places… but legislators like Joe Biden can roam free across campuses and events extolling own ethical virtues… for a fee... often paid for by students' tuitions.

Friday, May 19, 2017

19/5/17: A Reminder: Social Security is Only Getting More Insolvent...


On foot of my earlier post on U.S. household debt, it is worth mentioning another, much-overlooked in the media, fact concerning U.S. real economic debt crisis. This fact is a staggering one, even though it has been published a year ago, back in April 2016.

Based on the 2016 OASDI Trustees Report, officially called "The 2016 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds" (see link here: https://www.ssa.gov/oact/TR/2016/index.html).
  • U.S. Social Security's total income will exceed total cost of Social Security payouts through 2019. However, beyond 2019, interest income and money taken out of reserves will have to cover the funds required to offset Social Security's annual deficits until 2034.
  • Assuming the U.S. Presidential Administrations and the Congress continue business as usual approach to Social Security, the federal government payroll taxes will only be able to cover roughly 75% of scheduled retirement benefits until 2090
  • As the result, the Social Security Administration now projects that unfunded obligations will reach USD 11.4 trillion by 2090 or some $700 billion higher than the USD 10.7 trillion shortfall projected a year ago
  • Worse:  on an "infinite horizon" basis (netting Social Security expected future liabilities from forecast revenues) Social Security will face a USD 32.1 trillion in unfunded liabilities by 2090, or staggering USD 6.3 trillion more than 2015 projection
Chart below plots forecast Social Security unfunded liabilities corresponding to each forecast year:


The above clearly shows that the Social Security 'stabilisation' achieved in 2014-2015 is now not only erased, but is set back to what appears to be a rapid acceleration in liabilities back to 2008-2014 trend.

Yes, Social Security is a system in which people pay in taxes for an 'allegedly' ringfenced program that is supposed to supplement retirement. No, Social Security is not a program that is actually contractually ringfenced to provide anything whatsoever to those who pay into it. Which, really, means that the default on Social Security is looming large for the millennials and subsequent generations. And this raises the issue of what will happen to pensions provision across the entire U.S. Currently, even public sector pensions (across states and municipalities) are facing severe uncertainty and, in an increasing number of cases, actual cuts. Which raises public reliance on Social Security just at the time that the Social Security system is facing higher threats of insolvency. 

Meanwhile, household debt situation is getting from bad to awful (see this post: http://trueeconomics.blogspot.com/2017/05/19517-us-household-debt-things-are-much.html). 

The status quo is a prescription for a social, economic and political disaster. No medals for guessing what the Congress is doing about it all.

19/5/17: U.S. Household Debt: Things are Much Worse Than Headlines Suggest


Those of you who follow this blog know that I am a severe/extreme contrarian when it comes to median investor perceptions of the severity of leverage risks. That is to say, mildly, that I do not like extremely high levels of debt exposures at the macroeconomic level (aggregate real economic debt, which includes non-financial corporations debt, household debt and government debt), at the financial system levels (banking debt), at the microeconomic (firm) level, and at the level of individual investors own exposure to leverage.

With this in mind, let me bring to you the latest fact about debt, the fact that rings multiple bells for me. According to the data from the U.S. Federal Reserve, household debt in the U.S. has, as of the end of 1Q 2017, exceeded pre-2008 peak levels and hit an all-time high by the end of March.

Let's crunch some numbers.

  • Total Household Debt in the U.S. stood at USD 12.725 trillion at the end of 1Q 2017, up on USD 12.576 trillion in 4Q 2016. Previous record, reached in 3Q 2008 was USD 12.675, while the pre-Global Financial Crisis average was USD 10.112 trillion.
  • During pre-crisis period, Mortgage Debt peaked at USD 9.294 trillion in 3Q 2008. In 1Q 2017 this figure remained below this peak levels at USD 8.627 trillion. As flimsy as house price valuations can be, this means that there is no 'hard' asset underlying the new debt peak. If anything, the new overall household debt mountain is written against something far less tangible than real estate.
  • Student loans are up on previous peak (4Q 2016 at USD 1.310 trillion) at USD 1.344 trillion, as consistent with continued growth in the student loans crisis in the U.S.
Chart below illustrates the trends for total household debt:

Another key trend in household debt relates to debt defaults and risks. Here too 1Q 2017 data is far from encouraging. Pre-Global Financial Crisis average delinquencies (120 days or more overdue loans and Severely Derogatory delinquencies) average 2.07 percent of total debt outstanding. In 1Q 2017, some 29 quarters of deleveraging later, the comparable percentage is 3.0 percent. This is bad. Worse, take together, all household debt that was in delinquency in 1Q 2017 was 4.8 percent, which is still above 4.56 percent average for pre-2008 period. 


While overall delinquencies are not quite at problematic levels, yet, we must keep in mind the underlying conditions in which these delinquencies are taking place. Prior to the onset of the Global Financial Crisis, interest rates environment was much less benign than it is today toward higher levels of debt exposures: debt origination costs (direct cash costs) and debt servicing costs (income charge from debt) were both higher back in the days of the pre-2008 boom. Today, both of these costs are lower. Which should have led to lower delinquencies. The fact that delinquencies still run above pre-2008 levels implies that we are witnessing poorer underlying household fundamentals against which the debt is written.

Sadly, you won;t read this view of the current debt and debt burden issues from the mainstream media and analysts.

Thursday, May 18, 2017

18/7/17: Greece in Recession. Again.



Per recent data release, Greece is now back in an official recession, with 1Q 2017 growth coming in at -0.1%, following 4Q 2016 contraction of 1.2%. Worse, on seasonally-adjusted basis, Greek economy tanked 0.5% in 1Q 2017. The news shaved off some 0.9 percentage terms from 2017 FY growth outlook by the Government (from 2.7% to 1.8%), with EU Commission May forecasting growth of 2.1% and the IMF April forecast of 2.15%, down from October forecast of 2.77%.


Greece has been hammered by a combination of severe fiscal contractions (austerity), rounds of botched debt restructuring, and extreme fiscal and economic policy uncertainty since 2010, having previously fallen into a deep recession starting with 2008. Structural problems with the economy and demographics come on top of this and, at this stage in the game, are secondary to the above-listed factors in terms of driving down the country growth.

In simple terms, this - already 10 years long - crisis is fully down to the dysfunctional European policy making.


In real terms, Greek economy is now down almost 3 percentage points on where it was at the end of 2000 and even if we are to assume that the economy expands 2.15% in 2017, as projected by the IMF, Greece will still end 2017 some 0.76 percentage points below where it was at the start of its tenure in the euro area.

Meanwhile, the 2.1-2.15% forecasts are likely to be optimistic. Past record shows that, so far, since the start of the crisis, IMF’s forecasts were woefully inadequate in terms of capturing the true extent of the crisis in Greece.


As chart above shows, with exception of just two forecasts’ vintages, covering same year estimates (not actual forward forecasts), all forecasts forward turned out to be optimistic compared to the outrun (thick grey line for April 2017).

Another feature of the more recent forecast is that 2017 IMF outlook for Greece factors in worse expectations for 2018-2021 growth than ALL previous forecasts:


The key driver for this disaster is the EU-imposed set of policies and the resulting policy and economic uncertainty. In fact, if we were to take the lower envelope of growth projections by the IMF - projections that were based on the Fund’s assumptions that the EU will live up to its commitments to accommodate significant debt relief for the Greek economy from around 2013 on, today’s Greek real GDP would have been around 20-21 percent higher than it currently stands.


All in, Greece has sustained absolute and total economic devastation at the hands of the EU and its institutions, including ESM, ECB and EFSF. Yes, structurally, the Greek economy is far from being sound. In fact, it is completely, comprehensively rotten to the core and requires deep reforms. But this fact is a mere back row of violins to the real drama played out by the Eurogroup, the ESM and the ECB. The nation with already woeful demographics has lived through sixteen lost years, going onto seventeenth. Several generations are either face permanently damaged prospects of future careers, or have to deal with demolished hopes for a dignified retirement from the current ones, and a couple of generations currently in lower and higher education are about to join them.

Tuesday, May 16, 2017

16/5/17: M&As and Investment Climate: 1Q data





As an illustration to the point made a few weeks ago (see http://trueeconomics.blogspot.com/2017/04/28417-vuca-markets.html)  here is the latest data on aggregate deals volumes and deal values for M&As





16/5/17: Navigating the Bubbling Up Investment Seas


Here are the slides from my presentation at the IPU Conference 2017 two weeks ago:





















16/5/17: Technology: Jobs Displacement v Enhancement


Technological innovation is driving revolutionary changes across the labour markets and more broadly, markets for human capital. These changes are structural, deep and accelerating, and, owing to their nature, are not yet sufficiently understood or researched.

One theoretically plausible aspect of the technological innovation in terms of human capital effects is the expected impact of technology on demand for (and therefore supply of) different occupations. For example, we know that technology can act as a complement to or a substitute for labour.

In the former case, we can expect advancement of technology to create more jobs that are closely linked to enhancing technological innovation, deployment and productivity. In other words, we can expect more geeks. And we can expect - given lags in education and training - that as demand for geeks rises, their wages will rise in the short run before falling rather rapidly in the longer term.

In the latter case, there is a bit less certain, however. Yes, technology’s primary objective is to lower costs of production and increase value added. As a result, it is going to displace vast numbers of workers who can be substituted for via technological innovation. However, not all substitutable workers are made of the same cloth and not all technological innovation is capable of achieving unambiguous returns on investment necessary to sustain it. Take, for example, an expensive robot that costs, say, USD 600.000 a pop, but can only replace 3 lower skilled workers in a laundromat, earning USD16,000 per annum. So with benefits etc factored in, the cost of these 3 workers will be around USD70,000 per annum. It makes absolutely zero sense to replace these workers with new tech at least any time before the tech systems become fully self-replicating and extremely cheap. So, for really lower skills distributions, we can expect that jobs displacement by technology is unlikely to materialise soon. But for mid-range wages, consistent with mid-range skills, there is a stronger case for jobs displacement.

All of which suggests that we are likely to see a U-shaped polarisation process arising when it comes to jobs distribution across the skills segments: higher wage segment rising in total share of employment, as complementarity effects drive jobs creation here; and the lower wage segment also rising in total employment, as robots-induced increase in value added across the economy translates into greater demand for low-skills jobs that cannot be efficiently displaced by technology, yet. In the middle, however, we are likely to witness a cratering of employment. Here, the workers are neither complementary to robots, nor are they earning low enough wages to make expensive robots non-viable as a replacement alternative for labour.

Interestingly, we are already witnessing this trend. In fact, we have been witnessing it since the early 1990s. For example, Harrigan, James and Reshef, Ariell and Toubal, Farid paper titled “The March of the Techies: Technology, Trade, and Job Polarization in France, 1994-2007”, published March 2016, by NBER (NBER Working Paper No. w22110: http://ssrn.com/abstract=2755382) looked into “employee-firm-level data on the entire private sector from 1994 to 2007” in France.

The authors “show that the labor market in France has polarised: employment shares of high and low wage occupations have grown, while middle wage occupations have shrunk.” So the story is consistent with an emerging U-shaped labour market response to technological innovation on the extensive margin (in headcount terms). And more, the authors also find that inside margin also polarised, as “…the share of hours worked in technology-related occupations ("techies") grew substantially, as did imports and exports.”

However, the authors also look at a deeper relationship between technology and jobs polarisation. In fact, they find that, causally, “polarisation occurred within firms”, but that effect was “…mostly due to changes in the composition of firms (between firms). [And] …firms with more techies in 2002 saw greater polarization, and grew faster, from 2002 to 2007. Offshoring reduced employment growth. Among blue-collar workers in manufacturing, importing caused skill upgrading while exporting caused skill downgrading.”