Beating the software jungle

Selecting the economics software of the future



 
 

Thomas Cool, Consultancy & Econometrics

Scheveningen, Holland
cool /at/ dataweb.nl, http://thomascool.eu


No 99-04-05



 
 

Summary

Currently there is a jungle of software for economics, for both professional and educational software, and including the supportive mathematics and statistics. A comparison of 1993 showed and now in 1999 shows again - at least to this author - that Mathematica is the most useful and promising software, both for its elegant language and its breadth of application. A problem with Mathematica is its current price of about $1500 for a professional licence. Part of the solution would be to separate the language and interface and the engine. Once the Mathematica language is adopted as the lingua franca of science software, for which there are no legal barriers, there can be competition in front ends, interpreters and compilers. Another part of the solution in the short term would be coherent and determined discussion of the economics community (software users and purchasing departments) with Wolfram Research Inc. (WRI), the makers of Mathematica. Also, as there might still be a (natural or lock in) monopoly, there could be regulatory action that creates a public service utility. WRI could name its price for becoming a public utility company, and we might see whether Mathematica users are willing to pay that.
 
 




Introduction



Economists use software, both in education and professionally, and, certainly since the appearance of Shapiro & Varian (1998) "Information Rules", at http://www.inforules.com/, they are aware of the network effects here.

Currently there still is a jungle of software for economics, both in professional and in educational software, and currently this jungle is still growing. Since economics relies on math and statistics, these subjects must be included too, and then the jungle is even larger and denser. Thus, we still are in the phase of free competition. However, network economics will set in at some moment, and likely result into a natural or lock in monopoly that requires some level of regulation. And it can be wise to choose now, while the picking is good, instead of later on, when random forces have resulted into a lock in product of inferior quality. Though wisdom suggests that doing nothing often is the best strategy, there may come a moment to fight this jungle, and if possible beat it.

As we may already have wondered whether we need 100 different books that explain the ISLM model, we may wonder a fortiori whether we do need 100 different software packages. There is a difference between books and software, since we are more keen on transferability.

How are we to choose ? This paper will not be extensive, and only express my analysis. Software tests have not been extensive, so conclusions reached here are tentative.

Currently in 1999, I look back at the developments since 1993. In 1993 I already had more than ten years of software experience, with Algol, Pascal, Fortran, TSP, to name a few. When discovering Mathematica in 1993, at http://www.wolfram.com, I embraced it, see Cool (1995). While my own work henceforth basically used Mathematica, I have kept an open (though budget limited) eye for other developments. The same period has shown a growth in the number of software alternatives. I however have not found convincing arguments to change, and my impression is that of an increasing jungle instead of a convergence to some common standard. My objective now is to take stock again, to make some observations, and to do some suggestions.

While the discussion is open minded, it turns out that it still centers around Mathematica. The reader should be aware that a lot of my work thus is with Mathematica, and I even sell application software for it, see Cool (1999), so that I may have a personal lock in bias. Please check whether I am still level-headed. Please be aware too, that I do not want cross relations with Wolfram Research Inc. (WRI), the only providers of Mathematica, the product that my work relies on. So when I suggest to differentiate and to lower the price of the product, to separate the Mathematica language from front end and engine, and perhaps cutting up the company, I may still be biased in trying to be friends.

Note: This paper will include links. These may not remain valid over time, but they are still useful.
 
 

Other authors



This paper is not the only effort at creating sense in the jungle. For example, there is the journal of economic education at http://www.indiana.edu/~econed/jee.htm that has an eye on software. For example there is the JISC at http://info.ex.ac.uk/public_html/sp/jisc1.html, there is the US Dept. of Education initiative for k12 educational software and other new technology at http://www.ed.gov/offices/OERI/ORAD/LTD/panel.html (secondary schools), or the California Clearinghouse at http://clearinghouse.k12.ca.us/c/@eycDWixdNO3dQ/interact.html, or the Educational Software Institute ESI Online. Or there is the discussion on wider consequences on the organisation of education as by http://www.aln.org/alnweb/journal/vol2_issue1/hanna.htm. We will see more references below when we concentrate on professional software and software for higher education.
 
 

PM: Computer advantages for education



Our focus is on professional software and higher education. Let us first mention that the computer can bring a lot of advantages for education. Some points are:

  1. Working environment:
    1. Professionals will work with software, and hence they have to learn to use this. Students will have to learn, not just to punch buttons, but to relate input and output to theory.
    2. Software can compensate people for computational and analytical deficiencies. Software will be the walking stick for the blind.
  2. Level of innovation:
    1. Software can copy old formats, with the computer used only for speed and accuracy.
    2. Or there can be new technology for wholly new ways of doing things. (Not entirely sure whether this is 'new': For example, the technology of computer games could be applied to learning. Click on Kanji symbols, earn credits when you hit the right ones, and then go on to the next level. Another example is the wireless computer, see http://www.thejournal.com/magazine/97/aug/trends.html.)
  3. Location and time:
    1. Software can be used for groups (in class), or as stand alone.
    2. The computer allows time and distance learning. A teacher in Hong Kong can have students in New York. With cheap teachers in India, rich American kids can learn more and more; or there is competition to all teachers from Super Teacher.
  4. Teacher support:
    1. The computer can be used for record keeping, scoring and examination.
    2. In the future a student can select courses like from a salad bar, and do an examination at any particular moment.
Obviously, books still are hugely important, and some exams will have to be in paper and in person. The point for education is that we want software to add value.

For professional use there obviously is the responsibility of the professional involved to make an optimal choice on the software. For educators, there is the added responsibility of the care for the educated. It hurts to see that kids have to use bad software when the marginal cost of copying good software is negligible.
 
 

What software ?



As the internet attracts all software initiatives, we may look on the internet.

Using http://www.altavista.com I did various searches on keywords such as economics, software, math, course, teaching, interactive, computer, criteria. For example "+math +course +interactive". This search verified the existence of a jungle. The situation generally also is that sites provide some links with those very keywords, so the search finds the keywords that it is looking for (e.g. http://lorenz.mur.csu.edu.au/education/software.html, http://www.acm.org).

For statistics, a consultancy agency for example as Timberlake (http://www.timberlake.co.uk) presents a list of their statistical software. But this list is touched by chaos, since SPSS seems lacking - and who knows what else. Also the list gives only short explanations, and mixes all kinds of categories. The multimedia product ActivStats (http://www.awl-ile.com/datadesk-activstats/activstats/index.html) is mixed with neural networks, which is a wholly different category. (Except that it runs on a computer too.)

For math, a search of "+math +course +interactive" gives 49020 hits. There is a page http://forum.swarthmore.edu/library/browse/static/ed_topic/ that provides some info; there are the Transactions, at http://gams.nist.gov/toms/Overview.html. A performance comparison of the major mathematical software is at http://www.informatik.uni-frankfurt.de/~stst/ncrunch.html, and it is relevant enough for us to copy the summary conclusion in an appendix below. This comparison of mathematical software however concentrates on software that is intended for professional use. This is different from teaching math. Here, for e.g. undergraduates, we might think of http://boxermath.com/mar3/index.html as a relevant product - that wasn't easy to find.

For IMF, Fed, etc. there is the software for the big simulation models, like GAMS. On the other hand, for education, software is primarily generated in the context of books.

  1. For economics in general:
    1. A search for "+economics +software" found 175500 web pages
    2. Of course there is CodEc at http://netec.mcc.ac.uk/~adnetec/CodEc/CodEc.html
    3. A catalogue is at http://www.ilrt.bris.ac.uk/ctiecon/catalog.htm. Introductory economics has 15 entries, micro-economics has 8 entries, macro-economics has 21 entries, and then there are the statistics and econometrics departments.
    4. We see developments like http://www.mathscript.com/InteractiveEconomics/Edgeworth.ms
  2. For finance:
    1. A search for "+finance +software" found 973477 web pages.
    2. There of course is personal finance software (that can be very educational), there is a proposal for a clearing house at http://apollo.gmu.edu/~jmilam/fipse/fipse.html, there is http://www.iafe.org/ and http://www.afajof.org/ and a product like http://www.learninginsights.com/.
    3. For a book like Bodie & Merton "Finance": Prentice Hall has a 'companion website' http://cw.prenhall.com/bodie/
  3. For operations research / management science:
    1. A search of " +operations +software" gives 549053 Web pages
    2. A search of " +operations +software +education" gives 87750 Web pages
    3. Of course there is INFORMS at http://www.informs.org/
    4. For a book like Krajewski & Ritzman "Operations management", Addison-Wesley has a CD ROM of Excel applications at http://hepg.awl.com/kr/om5/software/om5soft.htm
    5. I found after a while http://mundoe.maths.mu.OZ.AU/~worms/ which contains a software page, with for example a link to http://www.taylorii.com/.
    6. Interestingly, where software and the development of a new research subject go hand in hand, like in work flow management, a software site also provides one of the better discussions of the subject, such as with http://www.ultimus1.com.

Some observations


  1. For the internet search above, one can consider it a show of strength of Mathematica that it succeeded in being listed not only in math sites, but also in statistics and economics.
  2. The software comparison reproduced below, with speed removed as a criterion, shows that Mathematica still is strong as 'math software'. It is a win as a professional working environment if one includes the quality of the language and the application range, and if one is lenient on speed since this is easier solved by technology than design problems. In the following I however will say "Mathematica (or Excel etc.)" when useful to maintain generality.
  3. We find that textbook authors start providing their own software. It is obviously useful that software fits the text, the used symbols, the graphs etcetera. One would tend to say: Let a thousand flowers bloom. And: As there also are many different car types, let us simply enjoy the fruits of competition. But:
    1. To repeat: As we already wondered whether we need 100 different explanations of the ISLM model, we may wonder a fortiori whether we do need 100 different software packages ? For software, we are more keen to desire transferability. Why a model in Excel, and having to estimate it in TSP ? Why not have a single language - and a language that makes automated examination feasible ? (And this is not just ISLM, but also Edgeworth Boxes, LP, regression, ...)
    2. If author A has its 10 chapter book with software, and author B does so too, could we not want that A concentrates on chapter 1-5 and makes a better text and better software, and author B does the same for chapter 6-10 ? Can these authors (Stiglitz, Taylor, Colander, ...) only teach, and can they not talk to each other ? The cost of software development may be lower than the cost of co-ordination. Or customers pay anyway.
  4. Many software producers seem to think that their software subject warrants an independent development.
    1. Authors may think that they save their students money by providing a stand alone software product that does not rely on an additional platform. But this can be penny-wise-pound-foolish.
    2. Each single specialised topic may cause a separate software product and cause a listing in a software page. But an application package of Mathematica (or any other environment like Excel etc.), will not be listed separately. A fortiori, my own software for example that covers economics, finance, operations and statistics, is not mentioned at specialised sites of economics, finance, operations and statistics, since the idea seems to be that it suffices to link to Mathematica. (Posh perhaps: not to be posted as any ordinary program ...)
  5. A good product like ActivStats relies on the less know product Data Desk. Why can't we have a good multimedia shell around Mathematica (or Excel etc...), and why can the additional Data Desk features not be applications of Mathematica - and thus henceforth be used in other applications as well ? A good product like Taylor II for queueing and simulation relies, for special effects, on its own interpreter language that has crude features: why not Mathematica ?
  6. We should be aware of the K12 link. Higher education will benefit if high school already uses related software.

Expectation



Taking various considerations into account, I find myself constructing the following table of software that is promising for the future. Excel will remain a basic tool that few will want to do without. The future however is to Mathematica. ActivStats and perhaps Boxermath are specialised multimedia products that cater to a specified learning situation. If these products, or multimedia learning per se, are to prosper, they will however need to use Mathematica as the algebraic language and for interfacing with other products. Also Excel will meet a moment on which it has to reconsider its language.
 
 

Table 1: Review of promising software


 
  Math Statistics Economics and Finance OR
and
MS
Open User friendly Automated examination
Mathematica yes yes yes yes Math Link yes development potential
ActivStats - yes - - Data Desk yes ?
Boxermath yes - - - ? likely ?
Excel ? basic basic basic ? basic ?

 
 
 

Mathematica



The power of Mathematica derives from its language. The syntax and concepts are agreeable to a trained mind. The language also has a wide scope: algebra, graphics, numerics, sound, ...

Mathematica currently also has drawbacks. For example: It could do much better on data handling and databases. Word for Windows is a better editor, in terms of text editing and e.g. automated indexing; but Word of course is no scientific editor. For some simple applications, Excel is simply easier, and in Excel there is a 'recorder': which means that one can program by clicking on the menu (which feature Mathematica should develop too).

A drawback of Mathematica is its price, as a stand alone professional licence costs about $1500 and a student licence about $125. Bill Gates has been sued for less.

As each economist knows, the value of prices is subjective. At 10% interest, the $1500 is a rent of $150 per annum, and for the continuous services rendered this is a bargain. People spend more on holidays or a higher class car or whatever. Obviously, an expenditure decision is difficult to make if one cannot see the advantages of software.

Economists might ask Wolfram Research Inc., the makers of Mathematica, to reconsider its pricing policy. As Nicholas Economides writes at http://raven.stern.nyu.edu/networks/wsj.html: "A striking example of failure to implement the right pricing and licensing strategies is provided by Apple. Through the 80s, Apple sold its computers and Mac operating system at monopoly prices attempting to maximize short run profits with full knowledge that this strategy hurt market penetration and future profits. Similarly, Sony set high prices and refused to license its Beta system until it was too late. Both Apple and Sony were caught in the arrogant belief that, since their products were "the best," they could rightfully set prices at monopoly levels. Events proved them both wrong. VHS won through wide (and cheap) licensing of its technology. Microsoft's operating systems won by attracting independent applications software. If Apple had used the appropriate licensing and pricing strategies, the Mac OS could today have the lion's share of the PC operating systems market rather than Windows95."

One would qualify the latter analysis with the market power of IBM, so that the argument on Apple is not as clear-cut, and see the history of Atari. But the point certainly requires consideration, and is relevant for the current software situation. Indeed, a highly priced and little differentiated Mathematica limits the market, and a limited market causes developers, teachers and text book authors to look for other environments. An interesting suggestion of Shapiro and Varian is to differentiate the product in at least three classes (student, professional, three star super special), so that most customers select the middle class. An aspect is that margins for resellers are higher for the highly priced goods, so that they may not put in much effort for the student versions. Similarly, supply chains and margin cumulations for application products currently cause higher prices, and market restructuring like Digital River is trying to do can enlarge and enrich the market.

Currently, Wolfram Research Inc. (WRI) is a privately owned company, it seems basically owned by Stephen Wolfram himself. One might argue: Who could we better trust this responsibility to ? Others might argue: If we all adopt Mathematica, will Stephen not become the new Bill Gates ? (Sorry, Bill, some people think this is something to worry about.) Whatever is the case, many might feel more comfortable if Mathematica is owned by a public utility service, that sells its products at break even prices, calculated over the life cycle, and allows a (formal) say of users in developments.

To summarise the points considered up to here:

Having made this intermediate summary, we should observe that we are a bit vague about what we mean by "Mathematica". We should distinguish the language from the front end and engine.

Note that the release of version 3 of Mathematica over version 2 showed the strength of the language, that remained the same. The big improvement concerned the user interface. No doubt future improvements will show this pattern. A subtle point here is that the language is very important for programmers who create an application, while it is often less important to front end users who are satisfied with menu clicks. A comparison can be made with the situation of people living downstream, and we may note that it isn't always smart to neglect what is required upstream.

A basic observation is that the 'lingua franca of software' is either a common good or a natural monopoly that needs regulation. As the English language belongs to all of us, and as the Mathematical notation dy/dx belongs to all of us, one would wonder whether Mathematica's D[y[x], x] is proprietary. If we had to pay for using English, perhaps we would stop speaking it. A user would not mind which engine evaluates the language, as long as the language is applicable and transferable over applications. And as language is for communication, there are always more partners involved, and it is difficult to see it as proprietary. (Indeed, "Naf Naf" is not for communication, since nobody knows what it means, and thus you have to pay for it.)

As Jensen & Wirth (1975) codified Pascal, and as Fortran 77 was subjected to ANSI standard (http://web.ansi.org/), see Ashcroft et al. (1982), there was no claim for a single engine, and all kinds of compilers were created. I agree with Tullock (1989) that we should make 'idea' inventions more attractive (i.e. ideas that are easily copied need market protection), but the ideas of Stephen Wolfram cs. on the Mathematica language do not warrant the level of protection created by regarding it as a monopoly (or making! it into one). A search at the ANSI site for FORTRAN gave a link to the "Information Infrastructure Standards Panel", but a search for Mathematica did not give a single hit. Perhaps Mathematica still is a young language, but we may also wonder whether we would want ANSI involvement.

For some people at WRI these comments can be harsh. First you make something, and sweat while making it, then it becomes popular, and suddenly the people who you provided with a service start talking about monopolistic exploitation and stuff, and they threaten to take away your baby. A way for WRI to understand this, is to see that one really cannot sell a language, and that the sale of the product also involves a commitment of the buyers. But, the transition to a better situation need not require pain. WRI could name its price for becoming a public utility company, and we might see whether (prospective) Mathematica users and venture capitalists are willing to pay that.
 
 

Conclusions



The choice of the economics software of the future depends upon the network economics as discussed by Shapiro & Varian (1998) "Information rules".

We have taken stock of the software jungle, and though since 1993 many new software products have seen the day of light, Mathematica still is strongest and most promising. Given the widespread circulation of the Shapiro & Varian analysis, and given the obvious case of economics software selection, many people will reach the same conclusions, and this will enhance the position of Mathematica.

Given the quality of Mathematica the slow penetration in the market actually is a bit surprising. However, there are clear network economics causes for this. One obstacle in this process is the current pricing policy for Mathematica, that forsters the generation of competitor products. Wolfram Research Inc. should differentiate its product and allow for lower prices if it wants to remain alive.

Also, for all parties involved it would be best to separate the Mathematica Language from the Front End and the Kernel (engine). New companies can compete on better front ends, e.g. multimedia interfaces, and engines and application packages.

The process of adoption of the language as the lingua franca of scientific software would speed up if WRI would become a public utility service company. WRI, as the leading company in scientific computing, could take up this challenge, and involve its clients and near competitors in the changeover.

Of course, there are many examples in history where expectations and promises did not come true. For example, Bill Gates may buy up Gauss and turn it into 'Science Excel' or whatever, and Mathematica then will end up in the back yard of history. (Sorry again, Bill.)
 
 

References



Ashcroft et al. (1981), "Programming with FORTRAN 77", Granada

Cool (1995, 1996), " Economics progrms written in Mathematica", http://wueconb.wustl.edu/eprints/prog/papers/9508/9508001.abs

Cool (1999), "The Economics Pack. User Guide", ISBN 90-804774-1-9, http://thomascool.eu

Economides (February 25, 1998), "Letter to the Wall Street Journal on Path Dependence", http://raven.stern.nyu.edu/networks/wsj.html

Jensen & Wirth (1975), "Pascal. User manual and report", Springer

Shapiro & Varian (1998), "Information Rules", HBP, http://www.inforules.com/

Tullock (1989), "Changing incentives to make economics more relevant", ch 18, p235-247 in Colander & Coats (1989), "The spread of economic ideas", CUP
 
 

Appendix: Summary of a software review 1997

This summary has been taken from http://www.informatik.uni-frankfurt.de/~stst/ncrunch.html/. Please note that software tests can be dated because of the arrival of new versions. Also, one should select the weights of the rows while considering the application.

"The summary should set the results of the speed comparison, the functionality of the programming environment, the data import/export facilities and the availability for different platforms in relation to the results of the comparison of the functionality. The relation between these four tests is 39:2:5:8:46. "
 
 
 
Test
GAUSS
Mac-
syma
Maple
Mathe-
matica
Mat-lab
Mu-PAD
O-Matrix
Ox
S-Plus
(3.2.29)
(2.2.1)
(V4)
(3.0.1)
(5.1)
(1.3.0)
(3.2)
(1.11)
(V4)
Comparison of the Mathematical functionality (38%)
88.276%
51.608%
52.073%
78.455%
76.204%
22.338%
35.730%
41.330%
61.813%
Comparison of the graphical functionality (10%)
52.174%
47.826%
56.522%
78.261%
82.609%
30.435%
52.174%
17.391%
82.609%
Functionality of the programming environment (8%)
50.000%
36.667%
30.000%
60.000%
73.333%
40.000%
36.667%
40.000%
56.667%
Data import/export (5%)
38.462%
7.692%
7.692%
38.462%
23.077%
7.692%
11.538%
23.077%
69.231%
Available platforms (2%)
61.111%
33.333%
72.222%
77.778%
50.000%
83.333%
11.111%
50.000%
38.889%
Speed comparison (37%)
49.937%
10.971%
0.013%
7.674%
39.975%
1.658%
70.799%
66.207%
37.176%
Overall result
64.384%
32.437%
29.674%
48.757%
60.029%
17.397%
48.723%
47.295%
54.277%