Collateralization of Assets, Over-Extension of Credit, and Free Trade: An Empirical Analysis in Search of Justice and an Expanding Middle Class Part 3

Print pagePDF pageEmail page

Charalambakis

  • Dr. John E. Charalambakis is the Chief Economist at Blacksummit Financial Group, Inc. Lexington, Kentucky. He is also with the Adjunct Faculty at Patterson School of Diplomacy, University of Kentucky.

Coauthored by: Dr. David Coulliette of Asbury University and Dr. Kenneth Rietz of Centre College

 This article will be posted in three segments due its length. This is Part III.

The Empirical Results: Methodology, Selection of Variables, and of Countries

We need to emphasize in the beginning of this section that the results and analysis here is preliminary and research is already underway for better understanding of the ideas that have been developed in this paper. Since part of the research is to determine which factors in the four infrastructure areas would contribute to the emergence of a middle class (once capital formation has taken place via the means of international trade), the first (and admittedly most subjective) step was to develop an initial list of factors that describe the infrastructure for each country.

Data were collected from different sources, such as the World Bank, the IMF, different branches of the UN (UNDP, and UNCTAD), and the World Factbook for the 2005 year. Note that the first step of the algorithm scaled the data by dividing by the maximum absolute value that occurred for each data variable, to prevent the larger-scale factors (such as exports) from overwhelming the smaller-scale factors (such as student-teacher ratio) in the model. This allowed us to readily compare the effectiveness of the coefficients that we obtained.

Since part of the purpose was to establish a proof of concept for using the Support Vector Machine (SVM) and a broad scope of infrastructures, originally it was decided that limiting the number of countries would achieve the purpose. We selected small lists of countries under each of the three categories frontier, emerging, and developed. Frontier countries are ones that most economists would agree have very few people in the middle class, but have the potential, such as sufficient resources, to develop one. Emerging countries have a middle class that is growing. Developed countries have a mature and stable middle class. An initial run used a very small sample of countries in each category. It was exceedingly successful, prompting an expansion of the lists. This paper details the results of the newer results.

Discussion of Algorithm

The mathematical technique used to determine the significance of the factors was the Support Vector Machine (SVM).  This is a classification method from learning theory that uses a set of input training data {(x1, y1), (x2, y2), . . ., (xk, yk)} where the xi represents vectors of dimension n and the y values are assigned a value of +1 or -1, depending on whether the point is inside a set or not.  For the purposes of this study, the x vectors hold the list of factors that describe the state of a given country that may be characterized as having a middle class (y = +1) or not (y = -1).

In common with most learning algorithm, the SVM algorithm operates in two phases. In the training phase, the SVM model with a linear kernel takes the training data and produces a bias value b and a vector c (dimension n) of coefficients. The testing phase of SVM  is then run on all the data, using just the n-dimensional vectors even if the y-value is known. The algorithm will calculate the dot product of the coefficient vector with the data vector and then subtract off the bias value. If the resulting number is positive, the algorithm predicts a y-value of +1; if it is negative, the algorithm predicts a y-value of -1.  Running the algorithm on the training data verifies that the training phase worked well.

This particular implementation of the SVM algorithm uses what is termed a linear kernel. It was chosen because of the limited number of countries in the data set, and because the linear kernel tends to be the worst performer. If this kernel works, then increased data and a non-linear kernel should work much better. A non-linear kernel has an additional step when the data is transformed non-linearly before the coefficients are determined.

A non-linear kernel will be important as this study proceeds. An example will illustrate the reason. One factor that is critical to the development of the middle class is the extension of credit. People in the lower classes do not have the capital with which to start a business and thereby move to a higher class. The marvelous success of microloans illustrates this point. Therefore, in a linear model, the amount of credit extended would certainly have a positive coefficient. That would imply that as more credit is extended, even greater benefits accrue. But at some point, credit may become overextended, and become detrimental to the middle class (see concluding note.) This is arguably a significant factor in what caused the collapse of the middle classes of Argentina and Mexico in the recent past. That means that the extension of credit must, at that point, have a negative coefficient. Only a non-linear approach to modeling the middle class can accommodate both aspects of the extension of credit. Similar comments could be made about other data, such as inflation (CPI), which has a rather small range of values considered healthy, while values much outside that range are considered detrimental to a country’s economy.

Analysis of Data: Initial Run

First, we ran SVM on a collection of 44 countries using 45 factors for each country. The training set was the collection of 17 frontier countries and 10 developed countries. The testing phase consisted of finding the predictions for those as well as the 17 emerging countries. (The same process was used during the reduced factor run of SVM.) The results are summarized in Table 1 on the next page, giving the SVM output value, but not the prediction, which is easy to determine from the sign of the output.

The first 17 rows of the table list the frontier countries, the next 17 rows list the emerging countries, and the last 10 rows list the developed countries. There are also two columns. The first numeric column gives the SVM output using all 45 factors for training and testing. The other numeric column will be explained below.

The results show very clearly that the SVM algorithm (even with the linear kernel) works very well. The frontier countries, except for Thailand, all fall into the SVM output range of -1.0 to -1.6; the emerging countries mostly fall in the range from -0.8 to 0.0; the developed countries, except for South Korea, fall in the range 0.6 to 1.5. These results, especially for the frontier and developed countries, form a primary validation of the SVM algorithm; it does seem to be doing what we want it to do. The separation between the ranges for the different categories of countries also seems remarkably large, providing further evidence that the algorithm is working.

Country 45 Factors

 

10 factors
Albania

-0.9999

-1.0479

Angola

-1.5701

-1.3473

Bolivia

-1.0987

-1.2265

Ethiopia

-1.6106

-1.4777

Georgia

-1.2053

-1.1505

Ghana

-1.4369

-1.2124

Guatemala

-1.1876

-1.0855

Indonesia

-1.2804

-1.2320

Kazakhstan

-1.0000

-0.8798

Kenya

-1.4301

-1.3824

Lebanon

-1.0003

-1.0088

Morocco

-1.1062

-1.0533

Nigeria

-1.4359

-1.3758

Peru

-1.0963

-1.0486

Philippines

-1.1356

-1.1535

Thailand

-0.9608

-0.8136

Venezuela

-1.0001

-1.0002

Argentina

-0.6834

-0.7224

Botswana

-0.4742

-0.5624

Brazil

-0.7782

-0.9008

Chile

-0.4801

-0.3891

China

-0.1282

-0.6589

Czech Republic

0.4781

0.0827

Egypt

-1.1327

-1.1795

India

-1.1646

-1.1560

Iran

-1.1666

-1.0534

Malaysia

-0.2523

-0.4184

Mexico

-0.8208

-0.7842

Poland

-0.4681

-0.4366

Romania

-0.8441

-0.7429

Russia

0.0256

-0.6292

South Africa

-0.5709

-0.5240

Turkey

0.5125

0.8718

Ukraine

-0.5202

-0.7147

Australia

0.9445

1.0000

Canada

1.0754

1.0976

France

1.2936

1.0998

Germany

1.4943

1.5142

Japan

1.5026

1.0235

Republic of Korea

0.1209

-0.0290

Singapore

0.9340

0.9998

Sweden

1.0516

1.2535

United Kingdom

0.6674

1.0005

USA

1.4181

1.5219

Table 1

Reduction of Factors

It could easily be argued that with 45 factors and 44 countries, it is easy to expect results of this caliber. So, we attempted to reduce the number of factors used, still regarding the conceptual framework.

The factors used in the remainder of this discussion are as follows:

v For physical infrastructure:

  • Paved roads in kilometers per capita
  • Number of cell phones per capita

v  For social infrastructure:

  • Amount spent on healthcare per capita
  • Literacy Rate

v For financial infrastructure:

  • Private sector credit as a percent of GDP
  • GDP (PPP) per capita

v For legal infrastructure:

  • Corruption index (Transparency International)

v For international trade (in dollars):

  • Exports
  • Imports per capita
  • Foreign reserves per capita

Table 1 above lists the output of the SVM algorithm using only these ten factors, in the second numeric column. Table 2 below lists these ten factors, and the coefficients that the SVM algorithm generates for each. (It should also be noted that results equivalently good can be obtained with only six factors, showing that SVM is more than adequate for separating the categories of countries.)

Factor

Coefficient
Paved roads in km per capita

0.2416

Number of cell phones per capita

0.3675

Amount spent on healthcare per capita

0.7636

Literacy rate

0.05407

Private sector credit as a percent of GDP

0.1571

GDP (PPP) per capita

0.9074

Corruption index

0.8518

Exports (billions USD)

0.4490

Imports per capita

0.3133

Foreign reserves per capita

0.2819

Table 2

The following comments are in order: First, all developed countries, with the exception of South Korea, show up with SVM output values in the appropriate range. This exception appears puzzling at first glance, but an examination of the data shows that it is almost entirely due to a value of the corruption index that is considerably lower than for other developed countries.

Second, this time only the Czech Republic shows with a positive prediction, although Turkey is very nearly positive. This complies with the liberalization and openness that both countries have exhibited over the last two decades, both most likely the result of the incentive provided by the future possibility of membership in the European Union. We could then, make the claim that international openness and exchanges serve the purpose of forming capital and thus, advancing the formation of the needed infrastructures which in turn will lead to the creation of the middle class.

Third, the relatively weak positions of Egypt, India, and Iran need to be reviewed in a time series before any conclusion is reached.. However, it is also worth mentioning that just by trade alone China performs better in the SVM, a fact which by itself could help us understand a little better the value of international trade in forming the necessary cornerstones that a middle class needs.

Conclusion: A Word of Caution and Direction for Future Research

The empirical part of this paper should be viewed as a proof-of-concept attempt for using a multi-factor and linear approach to quantifying the extent to which international trade forms the basis of capital formation, which in turn advances the formation of infrastructures that create a middle class. These results seem to indicate that using international trade and the infrastructures as have been described above along with the SVM algorithm, is a feasible methodology, and is worth continuing in broadly the same direction.

However, at this point I would like very briefly to introduce the idea of what happens when things go to the extreme, especially when the financial sector’s interests diverge from the trade sector’s interests i.e. from the production or real economy’s interests.  When efforts are being made to sustain prosperity and the middle class with paper means rather than real assets and real production, then we will see a divergence of the production and real sectors interests from the financial sector’s interests.  The latter will tend to produce paper assets which will be over-collateralized, over-securitized, for the purpose of generating significant short-term profits. The table below shows the explosion of derivatives and other related instruments (CDOs, CLOs, etc.) in the last few years. It demonstrates the extent of irrational collateralization of “assets”, where the financial sector keeps pushing for more and more securitization of paper assets, which will be sliced into pieces and sold to individual and institutional investors.

Source: Bank of International Settlements, 2008

Of course, it seems that we are just start learning the lesson that these paper-assets are nothing more than paper, i.e. there is nothing behind them.  This is the phenomenon of extreme and irrational securitization and collateralization that is taking place in the U.S. and the EU, and which has been destroying the financial sector, because it can only create bubbles and bubbles usually burst. The bursting of the bubbles will create in turn instability not only in the economic sector but also in the political and social sectors, and therefore the whole economy’s cohesiveness may become unstable and questionable, which eventually may lead to significant destructions.  As direction for future research, it would be interesting to identify the possibility for economies to establish a rule by which they collateralize and securitize assets in a way that will not destabilize the economies.  The proposal for future research would be to form an index of internationalization of the economy – whether this is imports and exports as a fraction of GDP, foreign reserves, FDIs, currency swings, technology transfers, etc – and use this index as the compass/anchor of collateralization and securitization, so that the interest of the real economy (production) are not disassociated from the interests of financial capital, and thus do not jeopardize the sustainment of the middle class via misallocation of resources.

 References

Aristotle, 1985, Nicomachean Ethics. T. Irwin, trans., Hackett, IN

Bernstein William, 2008, A Splendid Exchange, Atlantic Monthly Press, NY

Chen, P.-H. & Lin, C.-J. & Scholkopf, B. A Tutorial on ?-Support Vector Machines, [available at  http://www.csie.ntu.edu.tw/~cjlin/papers/nusvmtutorial.pdf]

Crafts Nicholas, 2000, Globalization and Growth in the Twentieth Century, IMF Working Paper, march

Donaldson, James; Roberts, Alexander, ed., 1994, Ante-Nicene Fathers, The Teachings of the Twelve Apostles, Hendrickson Publishers, Vol. 7

Dunning, H. Ray, 1998, Reflecting the Divine Image: Christian Ethics in Wesleyan Perspective, Intervarsity Press

Edwrads Sebastian, 1998, Openess, Productivity and Growth: What Do we Really Know? The Economic Journal, Vol. 108, pp. 383-398

Goldberg P., and Pavcnik N., 2007, Distributional Effects of Globalization in Developing Countries, Journal of Economic Literature, Vol. XLV, No. 1

Hayek Friedrich, 1976, The Mirage of Social Justice, Vol. 2 of Law, legislation, and Liberty, Routledge and Kegan-Paul, London, UK

Harrison Ann and Hanson Gordon, 1999, Who Gains from Trade Reform? Some Remaining Puzzles, Journal of Development Economics, Vol. 59, No. 1

Hay Donald, 2001, The Post-1990 Brazilian Trade Liberalisation and the Performance of Large manufacturing Firms, Economic Journal, Vol. 111, pp. 620-641

Hobbes Thomas, 1926, Leviathan, Hafner, NY

Honan Park, 1983, Matthew Arnold: A Life, Harvard University Press, MA

Irwin, Douglas, 2002, Free Trade Under Fire, Princeton University press, NJ

Keller, Wolfgang, and Yeaple, 2003, Multinational Enterprises, International Trade, and Productivity Growth, NBER Working Papers, No. 9504

Lock John, 1983, The Second Treatise on Government, Hackett, IN

Mencius, The Mind of Mencius, D.C. Lau, trans., Penguin, NY

Mill, John Stuart, 1910, The Letters of John Stuart Mill, ed. Hugh Elliot, Longmans Green, NY

Montesquieu, 1989, The Spirit of the Laws, trans., Cohler, Miller and Stone, Cambridge University Press, NY

Nozick Robert, 1974, Anarchy State and Utopia, Basic Books, NY

Plato, 1982, The Republic, G.M.A.Grube, trans., Hackett, IN

Rawls John, 1971, A Theory of Justice, Harvard University Press, MA

Sachs Jeffrey and Warner Andrew, 1995, Economic Reform and the Process of Global Integration, Brookings Papers on Economic Activity, Vol. 25, No. 1

Sala-I-Martin, Xavier, 2006, The World Distribution of Income, The Quarterly Journal of Economics, Vol. CXXI, No. 2

Solomon Robert and Mark Murphy, 1990, What is Justice? Oxford University Press, NY

Vlastos Gregory, 1962, Justice and Equality, in R. Brandt, ed., Social Justice, Prentice Hall, NJ

Wood Adrian, 1999, Openness and wage Inequality in Developing Countries, in market Integration, Regionalism and the Global Economy, ed., R. Baldwin et al., Cambridge University press, NY

Zakaria Fareed, 2003, The Future of Freedom, Norton, NY

 

 

Leave a Reply

Your email address will not be published. Required fields are marked *