Constructing Epistemic Landscapes:
Methods of GIS-Based Mapping
Evers, Hans-Dieter and Genschick, Sven and Schraven, Benjamin
Center for Development Research (ZEF), University of Bonn, Center for Development Research (ZEF), University of Bonn, Center for Development Research (ZEF), University of Bonn
2 August 2009
Online at https://mpra.ub.uni-muenchen.de/17135/
MPRA Paper No. 17135, posted 06 Sep 2009 19:16 UTC
ZEF
W orking Paper Series 4 4
Center for Development Research
Department of Political and Cultural Change
ISSN 1864- 6638
Zent rum für Ent wicklungsforschung Cent er for Development Research
Bonn 2009
Hans- Diet er Evers, Sven Genschick, Benjamin Schraven
Constructing Epistemic Landscapes:
M et hods of GIS- Based M apping
Cent er for Development Research, Universit y of Bonn
Edit ors: H.- D. Evers, Solvay Gerke, Pet er M ollinga, Conrad Schet t er
Authors’ address
Prof. Dr. Hans Dieter Evers
Center for Development Research (ZEF), University of Bonn, W alter- Flex- Str. 3
53113 Bonn, Germany
Tel. 0049 (0)228- 73 4909: Fax 0228- 731972 E- mail: hdevers@ uni- bonn.de
www.zef.de
Benjamin Schraven
Center for Development Research (ZEF), University of Bonn, W alter- Flex- Str. 3
53113 Bonn, Germany
Tel. 0049 (0)228- 73 1824: Fax 0228- 731972 E- mail: schraven@ uni- bonn.de
www.zef.de
Sven Genschick
Center for Development Research (ZEF), University of Bonn, W alter- Flex- Str. 3
53113 Bonn, Germany
Tel. 0049 (0)228- 73 4919: Fax 0228- 731972 E- mail: svengen@ uni- bonn.de
www.zef.de
Constructing Epistemic Landscapes:
M ethods of GIS- Based M apping
Hans- Diet er Evers, Sven Genschick, Benjamin Schraven Table of Contents
1. Introduct ion: Epist emic Landscapes and Knowledge M aps 2
2. Examples of Thematic M aps and Landscapes 4
2.1 Distribut ion of Knowledge Producing Organisations 4
2.2 Distribut ion of Knowledge Assets 7
2.3 Distribution of Students in 3D 7
3. The Spatial Distribut ion of Knowledge Attributes 9
3.1 GIS: linkage of space and content 9
3.2 Data Input Formats 9
3.3 Spatial Data Representation 10
3.4 Problems of Using t he Third Dimension 10
4. Results: Knowledge M aps or Epistemic Landscapes 10
4 .1 Knowledge M aps as Analytical Tools 1 0
4.2 Statistical M easures of Knowledge Clusters 11
5. Conclusion: The Use of Epistemic Landscapes 16
References 17
Abstract
The const ruct ion of knowledge maps, demonst rat ed in t his paper, is designed t o show t he epist emic landscape of cit ies, count ries or regions. Knowledge asset s, knowledge producing and disseminat ing organisat ions are referenced t o spat ial object s and int egrat ed int o GIS. They are furt her represent ed in t hemat ic maps and in 3- D perspect ive graphs. Special at t ent ion is given t o mapping and measuring knowledge clust ers. St at ist ical procedures t o measure t he degree of knowledge clust ering are discussed and ways are indicat ed t o compare and det ermine t he emergence of knowledge clust ers. We conclude t hat t he const ruct ion of knowledge maps showing t he complexit y of epist emic landscapes will enhance t he chances of government agencies, companies and civic organisat ions t o underst and and use knowledge for development . This paper is in t he first place meant as guideline for t he relat ed analysis.
Key W ords
Knowledge and development , knowledge maps, epist emic landscapes, knowledge clust ers, Geographic Informat ion Syst em (GIS)
1. Introduction: Epistemic Landscapes and Knowledge M aps
The World Development Report of 1999 has drawn at t ent ion t o t he fact t hat knowledge is not evenly dist ribut ed wit hin count ries, regions or urban areas. This has been described as t he exist ence of a
“knowledge gap” or, in relat ion t o t he ICT backbone of a knowledge syst em, as t he “digit al divide”. It is t hen proposed t o close t hese gaps by appropriat e development policies. So far most st udies have t ried t o show t he narrowing or widening of knowledge gaps by using indicat ors, as provided by t he KAM (knowledge assessment met hodology) dat a base of t he World Bank Inst it ut e, like number of researchers per million populat ion, invest ment in R& D (research and development ) as percent age of GDP and ot her indicat ors. In most cases t hese dat a have measured knowledge gaps bet ween count ries or regions, wit hout paying closer at t ent ion t o t he geographical dist ribut ion of knowledge asset s as well as t he exist ence of knowledge gaps wit hin count ries, provinces or cit ies. We int end t o close t his part icular
“knowledge gap” by providing a met hodology t o show and measure t he geographical dist ribut ion of knowledge relat ed asset s, people or organizat ions, which we have referred t o as “epist emic landscapes”.
This paper is t hus designed t o discuss t he met hodology of creat ing knowledge maps and analyzing epist emic landscapes. Dat a and maps from ongoing research in t he M ekong Delt a in Viet nam1 are used t o illust rat e various met hodological issues and t o show examples of maps and graphs. As discussed elsewhere (Evers 2008), knowledge landscapes are formed by knowledge clust ers, knowledge hubs and t he dist ribut ion of knowledge asset s. Following t he work of Port er on t he compet it ive advant age of nat ions (Port er 2003; Port er 1990) t here has been an upsurge of research and dat a collect ion on indust rial clust ers (Sölvell 2009). These st udies usually assume t hat modern indust rial clust ers are cent res of innovat ion and t herefore desirable. Our concept of knowledge landscapes is less value laden and int ends t o show t he spat ial dist ribut ion of knowledge. Which t ype of knowledge landscape opt imizes int ellect ual or indust rial out put remains a quest ion t o be decided by furt her empirical research.
We define epist emic landscapes in a geographical sense, i.e. we refer t o t he spat ial dist ribut ion of knowledge asset s wit hin a predefined region (Evers 2008). The concept is not yet st andard social science t erminology. We are using t he t erm “epist emic” in line wit h “epist emic cult ure”, t he cult ure of knowledge product ion, as coined by Karin Knorr- Cet ina (Knorr- Cet ina 1999). The t erm “epist emic” has been used in different cont ext s. One line of argument refers back t o Bacon and 18t h- cent ury 'encyclopaedism' and defines an epist emic landscape as depict ing a synt hesis of knowledge (Wernick 2006). In Weisberg and M uldoon’s st udy a single epist emic landscape corresponds t o t he research t opic t hat engages a group of scient ist s. Agent based modelling wit h Net Logo soft ware is used t o model t he changing epist emic landscape according t o research st rat egies of participat ing scient ist s (Weisberg and M uldoon 2007).
Concept ually dist ance rat her t han Euclidean dist ance is shown in graphs, similar t o t hose used t o illust rat e social net works. In our st udy we follow a different pat h and focus on t he development st rat egies of government s, st rat egic groups, firms, research inst it ut es and t heir success in shaping t he epist emic landscape of a region. The allocat ion of human and financial resources creat es knowledge asset s which are geographically dist ribut ed and can be measured, mapped and made t o depict t he cont ours of an epist emic landscape.
Epist emic landscapes develop over long periods of t ime. They are seldom shaped by individual act ors, but more oft en by t he collect ive act ion of st rat egic groups (Evers and Benedikt er 2009; Evers and Gerke 2009). Firms connect ed by a common int erest t o capit alize on t he compet it ive advant age of clust ering have an impact on epist emic landscapes t hrough t heir locat ion decisions. M ore over government st rat egies t o develop knowledge- based societ ies and economies have oft en been decisive in shaping epist emic landscapes. Relevant development policies have been assessed in det ail elsewhere for M alaysia and Indonesia (Evers 2003), Singapore and Germany (Hornidge 2007). In fact , developing regions of high-
1 This refers t o ongoing research on knowledge management and knowledge governance in t he wat er sect or of t he M ekong Delt a (WISDOM project ht t p://www.zef.de/1052.0.ht ml), carried out joint ly by t he Cent er for Development Research (ZEF), Universit y of Bonn, t he Sout hern Inst it ut e of Social Sciences, HCM C and The M ekong Development Research Inst it ut e of Can Tho Universit y. Useful comment s by St effen Gebhardt , DLR, are grat efully acknowledged.
t ech indust ries, clust ers or knowledge hubs are, by now, st andard pract ice in many regional planning depart ment s around t he world.
In order t o visualize and analyze epist emic landscape we have developed knowledge maps2. A knowledge map is t he cart ographic represent at ion of capt ured informat ion and relat ionships which enables t he efficient communicat ion and learning of knowledge by observers wit h differing backgrounds at mult iple levels of det ail (M eusburger 2000). The individual it ems of knowledge included in such a map can be t ext , st ories, graphics, models or numbers. M aps can also serve as links t o more det ailed knowledge sources ranging from t ext - based groupware document s t o dat abase schemas as well as point ers t o implicit knowledge (such as expert s).
The creat ion of knowledge maps is by now a st andard t ool of knowledge management in organizat ions.
“In t heir simplest forms, knowledge maps serve as guides t o t he t ype of knowledge held bot h inside and out side t he organizat ion. They serve as locat ors of t hat knowledge, and locat ors of people holding part icular knowledge. They give a visual overview of t he knowledge available t o t he organizat ion” (Foles 1998:14).
The t erm „mapping“ is, perhaps unfort unat ely, oft en used in t he sense of „list ing“, i.e. creat ing a list of it ems, like organisat ions, persons or asset s. “Yellow pages” list ing names of persons or firms and t heir knowledge asset s are (oft en used) t o make knowledge easily accessible in large organizat ions and bureaucracies. In t his paper “mapping” is, however, used in a cart ographic sense. A list of knowledge producing organizat ions, i.e. research inst it ut es and universit ies in Ho Chi M inh Cit y and t he M ekong Delt a3, list s next t o name and funct ion also t he address or t he geographical coordinat es. Furt hermore t he geographical dist ance t o ot her organisat ions is capt ured.
Figure 1: Example of Knowledge M apping / Listing Research Organizations in the M ekong Delta
No. English Translation of organization Latitude Longitude
Function 1 Cuu Long Delt a Rice Research Inst it ut e 10.104799 105.620077 wr 2 Sout hern Fruit Research Inst it ut e 10.406300 106.122190 wr 3 Biot echnology Research and Development Inst it ut e 10.028193 105.770270 wr 4 M ekong Delt a Development Research Inst it ut e 10.030410 105.766008 wr 5 Hoa An Bio- Diversit y - Applicat ion - Research Cent re 9.869289 105.774001 wr 6 Tri Tue Viet Humane Resource Dev. & Training Cent re 10.034732 105.779015 n 7 Comput er, Science & Technology Cent re 10.251991 105.971513 n 8 Research Cent re for Rural Development 10.037803 105.786273 wr 9 Research Cent re for Social Sciences and Humanit y 10.037803 105.786273 n 10 Cent re for Science Applicat ion and Technology Transfer 9.935052 106.342977 n 11 Research Cent re for Agricult ure and Rural Development 10.354048 106.371293 wr
Not e: wr=wat er relat ed research; n= no wat er relat ed research
These list s can easily be used t o build t hemat ic maps. These maps show t he dist ribut ion of dat a in geographical space. We shall, however, concent rat e on one part icular t ype of maps, showing t he
“clust ering” of organisat ions. M aps of t his kind are used in geographical or sociological research or creat ed for use in informat ion syst ems like t he WISDOM informat ion syst em, and for regional planning.
Clust er research goes back t o indust rial locat ion t heory and assumes t hat proximit y reduces t ransact ion cost and spurns innovat ion. Dat a banks, like t he European Clust er Observat ory (ht t p://www.clust erobservat ory.eu) or t he Viet namese Provincial Compet it ive Index
2 All maps in t his paper have been designed and produced by t he aut hors on t he basis of dat a generat ed wit hin t he WISDOM Project .
3 This list was compiled by Tat jana Bauer in 2008- 09 in t he course of her doct oral t hesis research wit hin t he WISDOM project .
(ht t p://www.pciviet nam.org), are designed t o ident ify indust rial clust ers wit hin nat ions. Our approach looks at knowledge clust ers wit hin cit ies, provinces and regions and uses t he exact geographical locat ion of organisat ions as a basis for analysis. Please not e t hat t his paper is designed as guideline for t he analysis of knowledge clust ers.
2. Examples of Thematic M aps and Landscapes
2.1 Distribution of Knowledge Producing Organisations
A map example is t he “dist ribut ion of organisat ions” in Ho Chi M inh Cit y (Evers and Bauer 2009). In t his case we have chosen t o work wit h dat a4 t hat have been geo- referenced via GPS and Google Eart h.
Aft er import ing dat a t o t he GIS (geographical informat ion syst em), t hey can be used t o creat e symbols in a simple map. A visual examinat ion of t he map shows t he unequal geographical dist ribut ion of organizat ions, but only wit h t he assist ance of st at ist ical calculat ions t he spat ial pat t erns can be furt her analyzed and compared.
Figure 2: Knowledge producing organisations within the district boundaries of Ho Chi M inh City.
By adding addit ional variables t he exist ence of know ledge clust ers can be analysed. As an example w e dist inguish bet ween wat er- relat ed and non- wat er- relat ed research inst it ut ions. By t his we indicat e, which inst it ut es are engaged in research on t he wat er sect or. The spat ial clust er analysis will give an indicat ion of agglomerat ion t endencies. Such an analysis makes sense if we follow t he assumpt ion t hat clust ers of research inst it ut es and knowledge- based firms are more product ive. The analysis can be
4 This dat a was compiled by Tat jana Bauer in 2008- 09 in t he course of her doct oral t hesis research wit hin t he WISDOM project
furt her enhanced by adding qualifying variables like for example “st aff numbers”, “publicat ions” and
“year of est ablishment ” et c.
In a first approach we used all “wat er- relat ed universit ies and colleges” t o analyse t heir st raight - dist ance (Euclidean) t o each ot her. The result is a “clust er map” (Figure 3) where each point is surrounded by a radius of 2000 met ers. The map finally represent s t he quant it y of overlaps of radii by a simple classificat ion. Thus t he differences bet ween high and low- densit y areas can be worked out very well. A more det ailed approach of clust ering spat ial dat a is given in t he following sect ion.
Figure 3: map of water- related Universities and Colleges in HCM C.
For a comparison of t he effect of government st rat egies on clust er format ion, maps of different regions can be visually examined. For a more sophist icat ed analysis st at ist ical measures have t o be employed which we are going t o discuss below. The following map (figure 4) shows t he low degree of organisat ional knowledge clust ering in t he M ekong Delt a of Viet nam.
Figure 4: M apping of knowledge producing organisations within M ekong Delta Provinces.
2.2 Distribution of Knowledge Assets
The t hird example of a map (Figure 5) uses t he dist ribut ion of Can Tho Universit y (CTU) st udent s wit hin t he M ekong Delt a provinces according t o t heir home province bet ween 1995 and 2008 as an indicat or for knowledge asset s. For Viet nam t his dat a is st at e- wide available. Besides an absolut e dist ribut ion of st udent s, t he map cont ains a gender classificat ion. This separat ion is displayed by pie chart s for each province.
Figure 5: CTU students according to their home provinces, regional and state- wide.
Finally t his map offers a predicat ion about t he degree of cent ralizat ion and surplus capacit y of Can Tho Universit y. Here it is possible t o dist inguish if t he universit y eit her has a regional or a nat ional relevance.
As shown in graph no 4 t he universit y t ends t o t he lat t er assumpt ion.
In t he long run more knowledge asset s could be added or combined so t hat we are able t o come t o a more comprehensive conclusion about t he dist ribut ion of “knowledge gaps” and “knowledge hubs”.
2.3 Distribution of St udents in 3D
In t he t erminology of knowledge management we of t en t alk about “landscapes”, “knowledge gaps” and
“knowledge hubs”. All t hese t erms are met aphorically t ouched.
Via t he illust rat ion of dat a by maps, we are able t o display t hose met aphoric t erms as graphical represent at ions. For t his purpose we creat e an art ificial landscape in a geographical sense in which t he surface charact erist ics are det ermined by t he respect ive dat a set . So, t he alt it ude of a given landscape element will be based on t he quant it y of a cert ain variable or at t ribut e.
Figure 6: Distribution of students, M ekong Delta and Ho Chi M inh City; data © General Statistics Office Vietnam, 2007
In our example we choose t o represent t he alt it ude of t he landscape by t he number of st udent s.
St udent s, graduat ed from a universit y or college, are here int erpret ed as knowledge asset s. By doing so, dat a peaks became (knowledge- ) hubs and dat a downs became (knowledge- ) gaps.
Finally it must be ment ioned t hat t here is lit t le addit ional informat ion benefit by using t he t hird dimension, but it is a vivid represent at ion of dat a t hat enhances t he underst anding of
complex issues. Three dimensional represent at ions are t herefore increasingly used in advert ising and in science cent res like t he recent ly opened knowledge clust er Fusionopolis in Singapore. In a lat er sect ion we shall discuss some problems of creat ing t hree- dimensional landscapes in great er det ail.
3. The Spatial Distribution of Knowledge Attributes
3.1 GIS: linkage of space and content
At t ribut es of dat a can be personal, inst it ut ional or regional and nat ional (et c.), so t hat dat a can always have a geo- reference. The linkage bet ween spat ial at t ribut es and cont ent makes it possible t o map t he dat a.
Working wit h a GIS means first of all t o st art t he management of geo- dat a. This includes t he st orage, t he t ransformat ion, t he st ruct uring, and t he creat ion of new geo- dat a, which at t ribut es, propert ies and ranges of values have t o be defined as well.
Before creat ing a map it is import ant t o always check whet her one can use an already exist ing map inst ead. Once t hat has been done t he dat a gat hering comes int o play. For t his, a det erminat ion bet ween primary, i.e. self- generat ed, and secondary, i.e. available dat a, is of import ance. Even t hough t he process of gat hering primary dat a usually is very t ime- consuming, it is more det ailed and represent s t he at t ribut es wit h much cont ext ual dept h. Secondary dat a on t he ot her hand are easier t o find and at lower prices t o get , even t hough it oft en lacks cont ext ual qualit y. Since t he gat hered dat a alone do not represent any kind of informat ion yet , t he gat hered dat a set t hen will have t o be t ransformed int o a meaningful map by using knowledge and t ransformat ion rules.
Because of t he vast amount of informat ion consist ing in t he primary as well as in t he secondary dat a set , one will have t o fulfil a reduct ion of complexit y, in order t o creat e t he aspired map. This can also include building dat a classes (before or aft er collect ing dat a).
3.2 Data Input Formats
Wit h t he except ion of t he geo- reference, t here are almost no ot her rest rict ions of dat a input . Geo- reference is t he st rongest requirement and st andard for mapping any kind of informat ion wit hin a GIS!
The scale levels of t he at t ribut es can vary. Their cont ent may cont ain nominal, ordinal, int erval or rat io scaled informat ion. Furt hermore dat a can be raised on different spat ial scales.
M ore examples of different scaled geo- references are: coordinat es, a cit y, a dist rict , a province, a st at e, a cont inent , (et c.). The different scaled geo- references are arranged in a hierarchical manner. An up- and downscaling of dat a depends on t heir resolut ion.
Anot her import ant guideline is t he usage of a coherent geographic and/ or reference project ion. If various shape files wit hin a map project are of different project ion, t heir visual display could be incorrect .
In case of t he simple purpose of adding addit ional informat ion t o already exist ing geo- object s t here are possibilit ies of import ing dat a wit hout a manual ent ry. For example, t o merge a t able wit h dat a (i.e.:
excel, dBase) t o already available at t ribut es, one just has t o set up ident ical t able st ruct ures (keys). This funct ionalit y increases t he collect ion of dat a by saving a lot of t ime.
Anot her useful applicat ion is depict ed by t he import of GPS- coordinat es. For inst ance ArcM ap5 offers such a feat ure for an easy and aut omat ic int egrat ion and display in shape- file6 format . By opening and
5 ArcM ap is component of ESRI´s ArcGIS. It is t he cent ral applicat ion for t he creat ion of maps. For t his approach ArcGIS 9.2 was used.
6A shapefile stores nontopological geometry and attribute information for the spatial
features in a data set. It consists of a main file, an index file, and a dBASE table. (ESRI 1997)
edit ing t he shapefile´s at t ribut e t able again a quick alt erat ion of numeric and/or t ext ual dat a can be done wit h ease.
3.3 Spatial Data Representation
As st at ed above, t he represent at ion of dat a is realised by point s, lines, areas or volumes. Furt hermore t able- and t ext - based dat a can be added. For a more det ailed illust rat ion classificat ion of dat a should be undert aken. By doing so, t he core st at ement s of t he dat a set move int o focus and more precise conclusions can be drawn.
Alongside t he represent at ion of dat a via “simple” or “nest ed” queries, which are charact erized as direct request s wit hout modifying dat a values by showing single or logical combined at t ribut es, t he dat a also can be manipulat ed by creat ing new geo- dat a from t he exist ing geomet rics.
For t he last ment ioned approach, t he clust er analysis of our wat er- or non- wat er relat ed organizat ions can be used as an example.
3.4 Problems of Using the Third Dimension
For t he creat ion of t he t hree- dimensional map (sect ion 3.4 above) we used dat a, which have been collect ed on a provincial scale, such as t he number of st udent s in Can Tho Cit y. Obviously st udent s are not dist ribut ed equally across provinces, but are concent rat ed in cert ain heavily populat ed areas. In fact we do not have any informat ion about how st udent s are dist ribut ed wit hin t his spat ial reference unit . In our map it is obvious t hat we carried out a more or less random int erpolat ion wit hin t his given reference. This problem is common t o most t hemat ic maps, but from t he st andpoint of st at ist ics t his approach is, st rict ly speaking wrong because we did an assessment wit hout any references of how t he dat a could be dist ribut ed in space. In t he long run we hope t o find a solut ion, which enable us t o int erpolat e t he dat a wit hin t he spat ial reference of collect ed dat a. The quest ion st ill remains: How t o handle t he space, for which dat a are not available?
4. Results: Knowledge M aps or Epistemic Landscapes
4.1 Knowledge M aps as Analytical Tools
So far we have shown how knowledge relat ed dat a set s can be visualized in t hemat ic maps. They have one t hing in common: t hey all use space or dist ance as an addit ional variable or, t o put it different ly, t o geo- reference dat a on knowledge product ion, knowledge asset s and knowledge flows. Dat a set s and maps all represent aspect s of what we have t ermed “epist emic landscapes”. The t erm “map” is used in t he st rict lit eral geographical sense. Our knowledge maps represent real geographical knowledge landscapes. Knowledge maps offer broad informat ion about t he spat ial dist ribut ion of knowledge asset s and knowledge product ion.
So far we have used knowledge maps mainly t o st udy t he process of clust ering. The underlying hypot hesis is, as explained elsewhere (Evers 2008), t hat a clust er of firms and organisat ions producing, t ransmit t ing or using knowledge enhances innovat ion and product ivit y and t hus provides a compet it ive advant age over ot her regions t hat show a lower degree of knowledge clust ering. Knowledge maps visualise t he degree of clust ering. We have t ried t o enhance t he “visibilit y” of knowledge by using t hree- dimensional maps or even animat ions. The didact ic value appears t o be evident but t he analysis has t o go beyond visual impressions. We have t herefore t ried t o use various st at ist ical t ools t o measure t he degree of clust ering, or in ot her words t o t est t he qualit y of epist emic landscapes and asses how t hey may cont ribut e t o social and economic development .
4.2 Statistical M easures of Knowledge Clust ers
An adequat e indicat or for t he spat ial densit y of a clust er will have t o be developed t o describe t he
“qualit y” of a knowledge clust er. In t he following some select ed approaches shall be int roduced. The densit y measures oft en used in t he nat ural sciences (e.g. t he st and densit y index SDI used oft en in forest ry (Reineke 1933), which basically work according t o t he funct ion “element s per unit of area/space” are not really applicable for t he measurement of spat ial densit y of knowledge clust ers. In t he case of knowledge clust ers, it obviously makes much more sense t o use t he int ernal dist ances wit hin a clust er for t he const ruct ion of a densit y measure.
One imaginable and very precise approach of doing t his is based on t he Euclidean or linear dist ance of each knowledge- producing organizat ion t o any ot her knowledge producing organizat ions wit hin a knowledge clust er. Therefore, a dat a mat rix wit h all coordinat es of t he involved organizat ions on bot h t he x- axis and t he y- axis has t o be generat ed. In t he mat rix t he Euclidean dist ances have t o be calculat ed whereas t he dist ances of t he organizat ions t o t hemselves have t o be excluded.
Table 2 Example of a dat a mat rix for t he calculat ion of t he Euclidian dist ances bet ween select ed knowledge producing organizat ions of a knowledge clust er
Name of institute
An Giang Universit as
Bac Lieu Universit y
Can Tho Universit y
Dong Thap Universit y …
Latitude 547768 579406 584535 568901 …
Name of institute
Latit
ude Longitude 1147099 1026762 1108858 1156715 …
An Giang Universit as
5477
68 1147099 124427 53049 23218 …
Bac Lieu Universit y
5794
06 1026762 124427 82256 130377 …
Can Tho Universit y
5845
35 1108858 53049 82256 50346
Dong Thap Universit y
5689
01 1156715 23218 130377 50346 …
… … … … … … … …
Table 2 shows t he dat a mat rix t hat needs t o be const ruct ed for t he det erminat ion of all Euclidean dist ance bet ween t he organizat ions of a knowledge clust er. The t ot al number of dist ance values, which is generat ed by t he mat rix, is always det ermined by t he equat ion n²- n, whereas n is t he t ot al number of organizat ions. Wit h t he generat ed values a new variable can be const ruct ed whereby all in all one variable per knowledge producing unit are produced. These variables cont ain all dist ances of t he respect ive unit t o all ot her unit s of t he knowledge clust er. Wit h t he mean value (and addit ionally t he st andard deviat ion, st andard deviat ion of t he mean, variance, range et c.) a coefficient s can be calculat ed t o measure t he spat ial densit y of a knowledge clust er. The disadvant age of t his approach is t hat t here is a high vulnerabilit y t owards out liers and ext reme cases. Pract ically, t his means t hat if a knowledge clust er has one organizat ion, which is locat ed in a great dist ance t o t he rest of t he clust er, t his clust er will have a very high value, especially for it s st andard deviat ion even if t he rest of t he clust er may be locat ed very close t oget her.
Figure 7: Spatial distribution of knowledge producing organisations in Ho Chi M inh
If we t ake t he example of research inst it ut ions in Ho Chi M inh Cit y (n=218) and t he M ekong Delt a (n= 42), t he average dist ance value for t he first ment ioned is 66.7 kilomet res and for t he last ment ioned 5.6 kilomet res wit h st andard deviat ions of t he mean of 5.6 kilomet res and 0.37 kilomet res. Principally problemat ic wit h t his approach is t he relat ive sensit ivit y for out liers and ext reme cases. This is not really t rue for t he example: t he Ho Chi M inh Cit y clust er has one inst it ut ion, which is locat ed in some nort h- west ern dist ance t o t he rest . But t his out lier has only some (soft ) impact on t he densit y score for t his clust er since wit hout t his inst it ut ion t he clust er would have a densit y score of 5.4 kilomet res in average.
To t ot ally overcome t his pot ent ial bias caused by out liers anot her t hird approach, which can be called t he “neighbourhood per radius approach”, can be applied. For t his met hod t he same dat a mat rix for t he calculat ion of t he Euclidean dist ances bet ween t he organizat ions of a knowledge clust er has t o be generat ed. The next st ep is t o check all t he dist ances whet her t hey are bigger or smaller t han a pre- defined dist ance value (e.g. 1.5 km, 2 km, et c.). If t hey are smaller, t he dist ance value has t o be t ransformed int o t he value 1, if not it has t o be t ransformed int o t he value 0 (see t able 3 for t he Ho Chi M inh Cit y clust er). If t hen t he sum per column or per row is count ed, t he out come is t he number of neighbouring organizat ions wit hin t he pre- defined radius for every organizat ion of t he clust er. The predefined dist ance is not t ot ally arbit rary, as an ideal dist ance for face- face communicat ion can be assumed or calculat ed. Net work dat a could be a possible source for t his calculat ion, as we shall argue below. Wit h t he st at ist ical measures ment ioned above coefficient s measuring t he spat ial densit y of t he respect ive knowledge clust er can be const ruct ed (for inst ance t he average number of neighbours per inst it ut e or t he st andard deviat ion of neighbours per inst it ut e).
Table 3 Example for t he “neighbourhood per radius approach” for t he value of 2 kilomet res
Name of institute
Academy of Polit ics - 2nd Base
Post s and Telecommuni- cat ions Inst it ut e of Technology
Viet nam Aviat ion Academy
Nat ional Academy of Public
Administ rat ion - HCM C Branch …
Latitude 695393 685946 682968 683799 …
Name of
institute Latitude Longitude 1199990 1193259 1193970 1191896
… Academy of
Polit ics - 2nd Base
695393 1199990
0 1 0 …
Post s and Telecommuni- cat ions Inst it ut e of Technology
685946 1193259
0 0 0 …
Viet nam Aviat ion Academy
682968 1193970
1 0 0
Nat ional Academy of Public
Administ rat ion - HCM C Branch
683799 1191896
0 0 0 …
… … … … … … … …
Cont rary t o t he first described approach, t he higher t hese values are t he denser is t he clust er spat ially.
The disadvant age of t his approach is t hat t he value for t he respect ive radius has t o be chosen purposely in order t o get sound result s. The radius size is severely det ermining t he result s for t he densit y score what can easily be demonst rat ed by t aking t he example of t he non- wat er research clust er of t he M ekong Delt a: t he score value (=t he mean number of neighbours) is 0.04 wit h a radius of r=1 kilomet re and is 2.6 for r=2 kilomet res (for Ho Chi M inh Cit y t he value changes from 13.6 for r=1 kilomet re t o 39.7 for r=2 kilomet res). A solut ion for t hat could be – if exist ing - t he analysis of respect ive social net work dat a, which for inst ance indicat e how oft en a member of one knowledge producing unit communicat es face t o face wit h a member of anot her unit . Based on t hat , a crit ical maximum dist ance for direct face t o face communicat ion wit hin a knowledge clust er can be est imat ed. Anot her disadvant age is t hat dist ances are only indirect ly int egrat ed int o t his measure. What in t he first int roduced approach has a very high impact on t he score values has a very low impact on t he accordant measure in t his approach.
A “compromise” posit ion bet ween t hese t wo approaches can be a t hird met hod, which can be called
“nearest neighbour”- approach – based on t he relat ed clust er analysis met hod. For t his approach, an Euclidean dist ance mat rix (as t he one of t able 2) is used and only t he lowest or minimal value for each knowledge- producing unit is t aken int o considerat ion. Accordingly, t he accordant values show t he dist ance t o t he nearest neighbouring unit for every observat ion. The so est ablished dist ribut ion wit h it s st at ist ical at t ribut es (mean value, st andard deviat ion, median, et c.) can be used t o describe t he spat ial densit y of t he knowledge clust er. For t he Ho Chi M inh Cit y clust er t he mean minimal dist ance has a value of 0.4 kilomet res whereas t he value for t he M ekong Delt a is 4.7 kilomet res. Generally, t he sensit ivit y t owards out liers and ext reme cases cannot be reduced wit h t his approach complet ely.
Anot her fourt h met hod, which st rongly mit igat es t he impact of out liers, is based on t he nat ural logarit hm. The values of t he Euclidean dist ance mat rix simply have t o be t ransformed wit h t he nat ural logarit hm funct ion, whereby t he impact of t he out liers and ext reme values will be st rict ly mit igat ed (see
t able 4). The mean for all average values per knowledge producing unit can be t aken t o const ruct a densit y measure for t his approach. The respect ive densit y score value for t he knowledge clust er in t he M ekong Delt a would be 10.7 whereas t he value for t he Ho Chi M inh Cit y clust er would have a value of 8.3. A disadvant age of t his approach is t hat t he dist ances bet ween t wo clust ers wit h very different measures of spread are not adequat ely reflect ed - as in t he example of Ho Chi M inh Cit y and t he M ekong Delt a where t he accordant score values only differ by t he fact or of 1.3.
Table 4 Example for t he logarit hmic t ransformat ion approach Name of
institute
An Giang Universit as
Bac Lieu Universit y
Can Tho Universit y
Dong Thap Universit y …
Latitude 547768 579406 584535 568901 …
Name of
institute Latitude Longitude 1147099 1026762 1108858 1156715
… An Giang
Universit as 547768 1147099
11,7314705 10,8789695 10,0526791 …
Bac Lieu
Universit y 579406 1026762
11,7314705 11,3175924 11,7781848 …
Can Tho
Universit y 584535 1108858
10,8789695 11,3175924 10,8266734
Dong Thap
Universit y 568901 1156715
10,0526791 11,7781848 10,8266734 …
… … … … … … … …
A last and fift h met hod, which t akes all possible dist ances wit hin a clust er int o considerat ion and at t he same t ime mit igat es t he impact of out liers, is t he normal dist ribut ion probabilit y index. In t his approach t he lat it ude and longit ude variables are used. Wit h t he means and t he st andard deviat ions of bot h variables as well as t he correlat ion coefficient bet ween t he lat it ude and longit ude variables t he probabilit y t hat t he accordant point is part of a normal dist ribut ion in t he t wo- dimensional vect or space is calculat ed. The probabilit y values are t hen mult iplied wit h t he dist ance bet ween t he lat it ude and t he longit ude values t o t he accordant mean values (see below for t he exact equat ions). The mean value of t he so creat ed values is t hen used as a densit y score, which is sensit ive only t owards very ext reme out liers. The accordant value for Ho Chi M inh Cit y is 973.1 and for t he M ekong Delt a it is 15268.3.
All in all, we have t hus proposed t he following five measures of spat ial or dist ance densit y:
1. The “M ean Euclidean Dist ance Index” (EDI) 2. The “Neighbourhood per Radius Index” (NRI) 3. The “Nearest Neighbour Index” (NNI)
4. The “M ean Logarit hmically Transformed Euclidean Dist ance Index” (LTEDI) 5. The “Normal Dist ribut ion Probabilit y Index” (NDPI)
M easure of densit y / clust ering index
Value for HCM C Value for M ekong Delt a
Equat ion (for
d
ij=Euclidean dist ance bet weenx
i andx
j)1. EDI 5.6 66.7
∑
=
i
a
iA 1 n
whereas
∑
−
≠=
i j
ij
i
d
a n
1 1
2. NRI (for r=1 km)
13.6 0.04
∑
=
i r i
r
b
B n 1
whereas
{ ∑ }
≤
∧
≠
−
⏐=
r d i j j
ij i
r i
ij
k d
b 1
1
3. NNI 0.4 4.7
∑
=
i
c
iC n 1
whereas
i
c
j= min
≠1
d
ij4. LTEDI 8.3 10.7
∑
=
i
ld
iD 1 n
whereas
ln( )
1
1 ∑
−
≠=
i j
ij
i
d
ld n
5. NDPI 973.1 15268.3.
)) (
² ( Pr
1 ∑ * ≥ −
−=
−i X i i
X X N ob n d
E
whereas N²=bivariat e normal dist ribut ion and =Euclidean dist ance of t o t he clust er mean
−
X i
d
x
i5. Conclusion: The Use of Epistemic Landscapes
In t his paper we have demonst rat ed how knowledge at t ribut es of various kinds can be made visible in a knowledge map. Knowledge is not evenly dist ribut ed in space but shows discont inuit ies, gaps and height s. We use t he concept of “epist emic landscape” t o allude t o t he complex t apest ry of knowledge asset s, inst it ut ions, organisat ions, hubs and clust ers. Since ant iquit y maps have been used t o provide orient at ion and direct ion. M oreover t hey “can bridge t he gap bet ween language and cult ure in t erms of communicat ing” (Hat field, 2006). M aps are t hus powerful t ools in aid of decision making, proper planning and resource allocat ion. They are put int o pract ice by indicat ing “how t o get from here t o t here”. They are in t hemselves reposit ories of knowledge, which can be ret rieved and put int o act ion.
Epist emic landscapes and knowledge maps are t herefore import ant part s of any informat ion syst em. By addit ionally comput ing st at ist ical measures of t he densit y of knowledge clust ers, comparat ive dat a can be int erpret ed. By correlat ing indicat ors, i.e. on knowledge out put (like research papers or pat ent s) wit h our knowledge densit y measure, t he effect iveness of knowledge clust ering can be st udied. So far a similar approach has been used t o st udy t he compet it ive advant age of regions, following t he earlier st udies of Port er and ot hers (Port er 1990; Sölvell 2009).
Last not least we refer t o t he didact ic value of maps. Informat ion and knowledge, in our case t he pot ent ial use of knowledge for development , have t o be dist ribut ed t o t hose t hat eit her benefit or suffer from t he applicat ion of knowledge for development . The concept of epist emic landscapes and maps showing t he complex and complicat ed business of producing and using knowledge enhance, we believe, t he chances of government agencies, companies and civic organisat ions t o underst and and use knowledge for development .
References
ESRI. 1997. “ESRI Shapefile Technical Descript ion. An ESRI Whit e Paper.
(ht t p://www.esri.com/library/whit epapers/pdfs/shapefile.pdf)
Evers, Hans- Diet er. 2003. "Transit ion t owards a Knowledge Societ y: M alaysia and Indonesia in Comparat ive Perspect ive." Comparat ive Sociology 2:355- 373.
—. 2008. "Knowledge Hubs and Knowledge Clust ers:Designing a Knowledge Archit ect ure for Development ." in ZEF Working Paper Series No 27. Bonn: Cent er for Development Research (ZEF), Universit y of Bonn.
Evers, Hans- Diet er , and Tat jana Bauer. 2009. "Emerging Epist emic Landscapes: Knowledge Clust ers in Ho Chi M inh Cit y and t he M ekong Delt a." ZEF Working Paper Series.
Evers, Hans- Diet er, and Simon Benedikt er. 2009. "St rat egic Group Format ion in t he M ekong Delt a - The Social Transformat ion of a M odern Hydraulic Societ y." ZEF Working Paper Series 35.
Evers, Hans- Diet er, and Solvay Gerke. 2009. "St rat egic Group Analysis." ZEF Working Paper Series 34.
Hornidge, Anna- Kat harina. 2007. Know ledge Societ y: Vision and Social Const ruct ion of Realit y in Germany and Singapore Berlin and London: LIT Verlag.
Knorr- Cet ina, Karin. 1999. Epist emic Cult ures: How t he Sciences M ake Know ledge. Cambridge: Harvard Universit y Press.
M eusburger, Pet er. 2000. "The spat ial concent rat ion of knowledge. Some t heoret ical considerat ions."
Erdkunde 54:352- 364.
Port er, M ichael 2003. "The Economic Performance of Regions." Regional St udies 37: 6- 7.
Port er, M ichael E. 1990. The Compet it ive Advant age of Nat ions. New York: The Free Press.
Reineke, L.H. 1933. Perfect ing a st and- densit y index for even- aged forest . Journal of Agricult ural Research 46:627- 638.
Sölvell, Örjan 2009. Clust ers – Balancing Evolut ionary and Const ruct ive Forces. St ockholm: Ivory Tower Publishers.
Weisberg, M ichael , and Ryan M uldoon. 2007. "Epist emic Landscapes and t he Division of Cognit ive Labor." in unpublished ms. Pit t sburgh. Penn: Universit y of Pennsylvania
Wernick, Andrew. 2006. "Comt e and t he encyclopedia " Theory, Cult ure & Societ y 23.
ZEF Development Studies
edit ed by Solvay Gerke and Hans- Diet er Evers
Cent er for Development Research (ZEF), Universit y of Bonn
Shahjahan H. Bhuiyan
Benef it s of Social Capit al. Urban Solid Wast e M anagement in Bangladesh Vol. 1, 2005, 288 S., 19.90 EUR, br. ISBN 3- 8258- 8382- 5
Veronika Fuest
Demand- orient ed Communit y Wat er Supply in Ghana. Policies, Pract ices and Out comes
Vol. 2, 2006, 160 p., 19.90 EUR, br. ISBN 3- 8258- 9669- 2
Anna- Kat harina Hornidge
Know ledge Societ y. Vision and Social Const ruct ion of Realit y in Germany and Singapore
Vol. 3, 2007, 200 S., 19.90 EUR, br. ISBN 978- 3- 8258- 0701- 6
Wolfram Laube
Changing Nat ural Resource Regimes in Nort hern Ghana. Act ors, St ruct ures and Inst it ut ions
Vol. 4, 2007, 392 p., 34.90 EUR, br. ISBN 978- 3- 8258- 0641- 5
Caleb R.L. Wall, Pet er P. M ollinga (Eds.) Fieldw ork in Dif f ucult Environment s.
M et hodology as Boundary Work in Development Research
Vol. 7, 2008, 192 p., 19.90 EUR, br. ISBN 978- 3- 8258- 1383- 3
Irit Eguavoen
The Polit ical Ecology of Household Wat er in Nort hern Ghana
Vol. 10, 2008, 328 p., 34.90 EUR, br. ISBN 978- 3- 8258- 1613- 1
Lirong Liu
Wirt schaf t liche Freiheit und Wachst um.
Eine int ernat ional vergleichende St udie Vol. 5, 2007, 200 p., 19.90 EUR, br. ISBN 978- 3- 8258- 0701- 6
Solvay Gerke, Hans- Diet er Evers, Anna- K.
Hornidge (Eds.)
The St rait s of M alacca. Know ledge and Diversit y
Vol. 8, 2008, 240 S., 29.90 EUR, br. ISBN 978- 3- 8258- 1383- 3
Charlot t e van der Schaaf
Inst it ut ional Change and Irrigat ion M anagement in Burkina Faso. Flow ing St ruct ures and Concret e St ruggles Vol. 11, 2008, 344 p., 34.90 EUR, br. ISBN 978- 3- 8258- 1624- 7
Phuc Xuan To
Forest Propert y in t he Viet namese Uplands. An Et hnography of Forest Relat ions in Three DaoVillages
Vol. 6, 2007, 296 p., 29.90 EUR, br. ISBN 978- 3- 8258- 0773- 3
Caleb Wall
Argorods of West ern Uzbekist an.
Know ledge Cont rol and Agricult ure in Khorezm
Vol. 9, 2008, 384 S., 29.90 EUR, br. ISBN 978- 3- 8258- 1426- 7
Nayeem Sult ana
The Bangladeshi Diaspora in Peninsular M alaysia. Organizat ional St ruct ure, Survival St rat egies and Net w orks
Vol. 12, 2008, 368 p., 34.90 EUR, br. ISBN 978- 3- 8258- 1629- 2
ZEF Working Paper Series, ISSN 1864- 6638 Depart ment of Polit ical and Cult ural Change
Cent er for Development Research, Universit y of Bonn
Edit ors: H.- D. Evers, Solvay Gerke, Pet er M ollinga, Conrad Schet t er
Nr. 1 Evers, Hans- Diet er and Solvay Gerke (2005). Closing t he Digit al Divide: Sout heast Asia’s Pat h Towards a Knowledge Societ y.
Nr. 2 Bhuiyan, Shajahan and Hans- Diet er Evers (2005). Social Capit al and Sust ainable Development : Theories and Concept s.
Nr. 3 Schet t er, Conrad (2005). Et hnicit y and t he Polit ical Reconst ruct ion of Afghanist an.
Nr. 4 Kassahun, Samson (2005). Social Capit al and Communit y Efficacy. In Poor Localit ies of Addis Ababa Et hiopia.
Nr. 5 Fuest , Veronika (2005). Policies, Pract ices and Out comes of Demand- orient ed Communit y Wat er Supply in Ghana: The Nat ional Communit y Wat er and Sanit at ion Programme 1994 – 2004.
Nr. 6 M enkhoff, Thomas and Hans- Diet er Evers (2005). St rat egic Groups in a Knowledge Societ y:
Knowledge Elit es as Drivers of Biot echnology Development in Singapore.
Nr. 7 M ollinga, Pet er P. (2005). The Wat er Resources Policy Process in India: Cent ralisat ion, Polarisat ion and New Demands on Governance.
Nr. 8 Evers, Hans- Diet er (2005). Wissen ist M acht : Expert en als St rat egische Gruppe.
Nr. 8a Evers, Hans- Diet er and Solvay Gerke (2005). Knowledge is Power: Expert s as St rat egic Group.
Nr. 9 Fuest , Veronika (2005). Part nerschaft , Pat ronage oder Pat ernalismus? Eine empirische Analyse der Praxis universit ärer Forschungskooperat ion mit Ent wicklungsländern.
Nr. 10 Laube, Wolfram (2005). Promise and Perils of Wat er Reform: Perspect ives from Nort hern Ghana.
Nr. 11 M ollinga, Pet er P. (2004). Sleeping wit h t he Enemy: Dichot omies and Polarisat ion in Indian Policy Debat es on t he Environment al and Social Effect s of Irrigat ion.
Nr. 12 Wall, Caleb (2006). Knowledge for Development : Local and Ext ernal Knowledge in Development Research.
Nr. 13 Laube, Wolfram and Eva Youkhana (2006). Cult ural, Socio- Economic and Polit ical Con- st raint s for Virt ual Wat er Trade: Perspect ives from t he Volt a Basin, West Africa.
Nr. 14 Hornidge, Anna- Kat harina (2006). Singapore: The Knowledge- Hub in t he St rait s of M alacca.
Nr. 15 Evers, Hans- Diet er and Caleb Wall (2006). Knowledge Loss: M anaging Local Knowledge in Rural Uzbekist an.
Nr. 16 Youkhana, Eva, Laut ze, J. and B. Barry (2006). Changing Int erfaces in Volt a Basin Wat er M anagement : Cust omary, Nat ional and Transboundary.
Nr. 17 Evers, Hans- Diet er and Solvay Gerke (2006). The St rat egic Import ance of t he St rait s of M alacca for World Trade and Regional Development .
Nr. 18 Hornidge, Anna- Kat harina (2006). Defining Knowledge in Germany and Singapore: Do t he Count ry- Specific Definit ions of Knowledge Converge?
Nr. 19 M ollinga, Pet er M . (2007). Wat er Policy – Wat er Polit ics: Social Engineering and St rat egic Act ion in Wat er Sect or Reform.
Nr. 20 Evers, Hans- Diet er and Anna- Kat harina Hornidge (2007). Knowledge Hubs Along t he St rait s of M alacca.
Nr. 21 Sult ana, Nayeem (2007). Trans- Nat ional Ident it ies, M odes of Net working and Int egrat ion in a M ult i- Cult ural Societ y. A St udy of M igrant Bangladeshis in Peninsular M alaysia.
Nr. 22 Yalcin, Resul and Pet er M . M ollinga (2007). Inst it ut ional Transformat ion in Uzbekist an’s Agricult ural and Wat er Resources Administ rat ion: The Creat ion of a New Bureaucracy.
Nr. 23 M enkhoff, T., Loh, P. H. M ., Chua, S. B., Evers, H.- D. and Chay Yue Wah (2007). Riau Veget ables for Singapore Consumers: A Collaborat ive Knowledge- Transfer Project Across t he St rait s of M alacca.
Nr. 24 Evers, Hans- Diet er and Solvay Gerke (2007). Social and Cult ural Dimensions of M arket Expansion.
Nr. 25 Obeng, G. Y., Evers, H.- D., Akuffo, F. O., Braimah, I. and A. Brew- Hammond (2007). Solar PV Rural Elect rificat ion and Energy- Povert y Assessment in Ghana: A Principal Component Analysis.
Nr. 26 Eguavoen, Irit ; E. Youkhana (2008). Small Towns Face Big Challenge. The M anagement of Piped Syst ems aft er t he Wat er Sect or Reform in Ghana.
Nr. 27 Evers, Hans- Diet er (2008). Knowledge Hubs and Knowledge Clust ers: Designing a Knowledge Archit ect ure for Development
Nr. 28 Ampomah, Ben Y., Adjei, B. and E. Youkhana (2008). The Transboundary Wat er Resources M anagement Regime of t he Volt a Basin.
Nr. 29 Saravanan.V.S.; M cDonald, Geoffrey T. and Pet er P. M ollinga (2008). Crit ical Review of Int egrat ed Wat er Resources M anagement : M oving Beyond Polarised Discourse.
Nr. 30 Laube, Wolfram; Awo, M art ha and Benjamin Schraven (2008). Errat ic Rains and Errat ic M arket s:
Environment al change, economic globalisat ion and t he expansion of shallow groundwat er irrigat ion in West Africa.
Nr. 31 M ollinga, Pet er P. (2008). For a Polit ical Sociology of Wat er Resources M anagement .
Nr. 32 Hauck, Jennifer; Youkhana, Eva (2008). Hist ories of wat er and fisheries management in Nort hern Ghana.
Nr. 33 M ollinga, Pet er P. (2008). The Rat ional Organisat ion of Dissent . Boundary concept s, boundary object s and boundary set t ings in t he int erdisciplinary st udy of nat ural resources management . Nr. 34 Evers, Hans- Diet er; Gerke, Solvay (2009). St rat egic Group Analysis.
Nr. 35 Evers, Hans- Diet er; Benedikt er, Simon (2009). St rat egic Group Format ion in t he M ekong Delt a - The Development of a M odern Hydraulic Societ y.
Nr. 36 Obeng, George Yaw; Evers, Hans- Diet er (2009). Solar PV Rural Elect rificat ion and Energy- Povert y: A Review and Concept ual Framework Wit h Reference t o Ghana.
Nr. 37 Scholt es, Fabian (2009). Analysing and explaining power in a capabilit y perspect ive.
Nr. 38 Eguavoen, Irit (2009). The Acquisit ion of Wat er St orage Facilit ies in t he Abay River Basin, Et hiopia.
Nr. 39 Hornidge, Anna- Kat harina; M ehmood Ul Hassan; M ollinga, Pet er P. (2009). ‘Follow t he Innovat ion’ – A joint experiment at ion and learning approach t o t ransdisciplinary innovat ion research.
Nr. 40 Scholt es, Fabian (2009). How does moral knowledge mat t er in development pract ice, and how can it be researched?
Nr. 41 Laube, Wolfram (2009). Creat ive Bureaucracy: Balancing power in irrigat ion administ rat ion in nort hern Ghana.
Nr. 42 Laube, Wolfram (2009). Changing t he Course of Hist ory? Implement ing wat er reforms in Ghana and Sout h Africa.
Nr. 43 Scholt es, Fabian (2009). St at us quo and prospect s of smallholders in t he Brazilian sugarcane and et hanol sect or: Lessons for development and povert y reduct ion.
Nr: 44 Evers, Hans- Diet er, Genschick, Sven, Schraven, Benjamin (2009). Const ruct ing Epist emic Landscapes: M et hods of GIS- Based M apping.