Koen Deconinck and Marijke Verpoorten, "Narrow and Scientific Replication of 'The Slave Trade and the Origins of Mistrust in Africa'", Journal of Applied Econometrics, Vol. 28, No. 1, 2013, pp. 166-169. Our dataset is stored both in ASCII format and in Stata's .dta format, which we used. In addition, we include our code (in Stata's do-file format). The full questionnaires and codebooks of the survey data used in our analysis can be found at www.afrobarometer.org. The file containing Stata data is NunnWantchekon_Dataset.dta. It is zipped in the file dv-stata-data.zip. There are two files containing raw data: NunnWantchekon_Dataset_label.raw and NunnWantchekon_Dataset_nolabel.raw. The paper uses survey data with qualitative answers which are coded numerically. In the "labelled" version, answers are written out in full, e.g. "not satisfied", "satisfied", and so on. In the "unlabelled" version, answers are converted to numerical codes. The do-file is NunnWantchekon.do. All three of these files are ASCII files in DOS format and are zipped in the file dv-files.zip. Unix/Linux users should use "unzip -a". The rest of this file documents the construction of the dataset used in our replication. 1. Narrow Replication: The narrow replication was performed using the original dataset used by the authors. This dataset is available in Stata's .dta-format from Nathan Nunn's personal website: http://www.economics.harvard.edu/faculty/nunn/data_nunn Nathan Nunn also provides other materials used in the construction of his dataset. In addition, the web appendix to the original paper explains the data sources used by Nunn and Wantchekon: http://www.economics.harvard.edu/faculty/nunn/files/Appendix_MS_AER_2009_0252_R2.pdf 2. Scientific Replication: For our scientific replication, we used the 2008 version of the Afrobarometer survey, instead of the 2005 survey used by Nunn and Wantchekon. The Afrobarometer survey data, questionnaires/codebook and other materials can be found at the Afrobarometer website (http://afrobarometer.org/) We followed the procedure outlined in the appendix to the original paper to construct our dataset in a comparable way. The construction of the dataset consisted of several steps. First, we linked ethnicities as reported in Afrobarometer 2008 to ethnicities as described by Murdock (1959)'s classification of African ethnicities. Second, for those ethnic groups which were not included in the original paper, we constructed ethnicity-based covariates similar to those used by Nunn and Wantchekon. We did this using documentation and extra data generously provided to us by Nathan Nunn. -- Some variables are based on historical/anthropological data (e.g. the Ethnographic Atlas), provided by Nunn. -- Other variables had to be calculated using Murdock's 1959 map of African ethnicities and their homelands (e.g. the size of the homelands and the distance to the coast) This calculation was performed using ArcGIS and using Nathan Nunn's digitized version of the Murdock 1959 map. Third, using extra information provided by Afrobarometer on the name of the town/village of every respondent in the survey, we used Google Maps to find the GPS coordinates of every respondent. With these GPS coordinates and the 1959 Murdock map, we used ArcGIS to find the ethnic group originally living in the current location of the respondent. Based on the variables collected in this way, it was straightforward to calculate all necessary covariates. Below, we provide more information on the ethnic groups included in our dataset. ETHIC GROUPS IN AFROBAROMETER 2008 Afrobarometer 2008 contains several ethnic groups not included in the Afrobarometer 2005. The extra ethnicities are mostly due to the addition of two new countries (Liberia and Burkina Faso) and the fact that in AB2005 the question about ethnicity had not been asked in Zimbabwe, while this question was included in AB2008. However, in several other countries the range of ethnicities in the survey has expanded as well. For example, in 2005 the Afrobarometer included only 6 ethnic groups for Ghana, one of which was "other northern languages". In the 2008 Afrobarometer, 24 ethnic groups are included for Ghana. This note details how many new ethnic groups are available in our replication dataset as compared to the original N&W dataset. In order to match Afrobarometer 2008 ethnicities with data on slave trades, we used the following procedure. First, we matched names of 2008 ethnicities with names of 2005 ethnicities, to exploit the correspondences made by Nunn and Wantchekon (2011) wherever available. If the ethnic group did not appear in the dataset used by Nunn and Wantchekon, we matched ethnic groups to Murdock's ethnicities using the following approach. First, if the name of the ethnic group as stated appeared in Murdock (1959), we matched it to a Murdock ethnicity. If the name, as stated, was not found in Murdock (1959), we used Olson (1996) to check alternative spellings, or we used information on the language spoken at home by the respondent (q3 in the Afrobarometer survey) combined with information from the Ethnologue: Languages of the World database (Lewis, 2009), available online at www.ethnologue.com, to find the corresponding ethnic group in Murdock (1959). Of the 27,713 individuals in the Afrobarometer Round 4 (2008), 3234 respondents stated their ethnicity in a way that inherently could not be matched to indigenous ethnic groups (Table 1). Table 1. European or Non-Indigenous Ethnic Identifications and Missing Values Description Number African 373 Afrikaaner 40 Afrikaans/Afrikaner/Boer 310 Amerivan or European 30 Coloured 230 Don't know 520 English 108 German 4 Missing 26 National ID only or don't think in those terms 850 Portuguese 9 Refused 58 Relacionado com o estado de espirito 8 Related to age 9 Related to class 4 Related to gender 15 Related to occupation 19 Related to political-partisan affiliation 3 Related to race 4 Related to regional origin (badio/sampadjudo) 33 Related to religion 10 White/European 16 Others 555 TOTAL 3234 Of the 555 individuals who gave "other" as their ethnic group, 306 also gave "other" as their home language. Of those who did indicate a specific language, many indicate a language which is linked to a specific ethnic group (in which case it would probably be incorrect to assign these individuals to that ethnic group, since they did not identify themselves with it) or a national language (e.g. Swahili), which makes it impossible to link them to an original ethnic group. For these reasons we have chosen to drop the group of "others" altogether. By matching names of the 2008 ethnicities with names of the 2005 ethnicities, we were able to link 181 ethnic groups (16,246 respondents) immediately and another 35 ethnicities (2,582 respondents) after adjusting minor changes in spelling (e.g. "khasonk" could be matched to "khassonke"). Eight extra ethnic groups (totaling 429 individuals) were matched with an ethnic group available in the N&W dataset but which N&W could not link to a Murdock name. Of the remaining ethnicities, the approach outlined above allowed us to find a match for 131 ethnic groups (counting 4,651 individuals). For 24 ethnicities (571 respondents) we were unable to find a corresponding Murdock name. Hence, in total we have -- 216 ethnic groups (18,828 individuals) corresponding to an ethnicity included in N&W and matched to a Murdock name -- 8 ethnic groups (429 individuals) corresponding to an ethnicity included in N&W for which they could not find a Murdock name -- 131 ethnic groups (4,651 individuals) not included in the original N&W dataset which we could match to a Murdock name -- 24 ethnic groups (571 individuals) not included in the original N&W dataset which we could not match to a Murdock name Ethnic groups on which covariates were needed: There are 69 Murdock ethnicities (accounting for 3,009 respondents) for which Nunn and Wantchekon's original dataset did provide slave trade data but for which we didn't have other covariates. The following table lists these ethnicities, for which we constructed the necessary covariates using the procedure outlined above. murdock_name | Freq. Percent Cum. -----------------------------+----------------------------------- ACHOLI | 179 5.95 5.95 ANYI | 19 0.63 6.58 ASSINI | 12 0.40 6.98 BANDA | 5 0.17 7.15 BASARI | 9 0.30 7.44 BASSA | 151 5.02 12.46 BENA | 18 0.60 13.06 BIRIFON | 9 0.30 13.36 BORANA | 7 0.23 13.59 BUA | 12 0.40 13.99 BUILSA | 6 0.20 14.19 BURA | 16 0.53 14.72 BUSANSI | 73 2.43 17.15 DAN | 80 2.66 19.81 DIULA | 6 0.20 20.01 GBANDE | 22 0.73 20.74 GOLA | 53 1.76 22.50 GREBO | 143 4.75 27.25 GRUNSHI | 35 1.16 28.41 GUANG | 10 0.33 28.75 GUIN | 12 0.40 29.15 GUN | 78 2.59 31.74 GURENSI | 89 2.96 34.70 GURMA | 49 1.63 36.32 HA | 48 1.60 37.92 HEIKUM | 8 0.27 38.19 HLENGWE | 30 1.00 39.18 IRAQW | 37 1.23 40.41 JUKUN | 7 0.23 40.64 KAMBERI | 10 0.33 40.98 KARABORO | 13 0.43 41.41 KINGA | 10 0.33 41.74 KISSI | 48 1.60 43.34 KONKOMBA | 11 0.37 43.70 KOREKORE | 87 2.89 46.59 KPELLE | 286 9.50 56.10 KRAN | 49 1.63 57.73 KRU | 70 2.33 60.05 LANGO | 187 6.21 66.27 LIPTAKO | 81 2.69 68.96 LOBI | 19 0.63 69.59 MADA | 9 0.30 69.89 MAMPRUSI | 16 0.53 70.42 MANYIKA | 80 2.66 73.08 MARGI | 12 0.40 73.48 MATENGO | 12 0.40 73.88 MATUMBI | 11 0.37 74.24 MENDE | 9 0.30 74.54 MUMUYE | 8 0.27 74.81 MWERA | 19 0.63 75.44 NGERE | 106 3.52 78.96 NGONYELU | 7 0.23 79.20 NUBIAN | 7 0.23 79.43 POGORO | 17 0.56 79.99 POPO | 6 0.20 80.19 RANGI | 11 0.37 80.56 RONGA | 25 0.83 81.39 SAPO | 11 0.37 81.75 SHAMBALA | 18 0.60 82.35 SHASHI | 27 0.90 83.25 TANGALE | 8 0.27 83.52 THLARU | 8 0.27 83.78 TOMA | 98 3.26 87.04 TSWANA | 7 0.23 87.27 TUKULOR | 296 9.84 97.11 VAI | 42 1.40 98.50 ZARAMO | 18 0.60 99.10 ZERMA | 2 0.07 99.17 ZIGULA | 25 0.83 100.00 -----------------------------+----------------------------------- Total | 3,009 100.00 References Lewis, M. Paul (ed.) (2009) Ethnologue: Languages of the World, 16th edition, SIL International (Dallas, Texas). Available online at www.ethnologue.com Murdock, G.P. (1959) Africa. Its Peoples and Their Culture History, McGraw-Hill Book Company (New York, Toronto, London), 456p. Nunn, N. and L. Wantchekon (2010) Web appendix for "The Slave Trades and the Origins of Mistrust in Africa", available at http://www.economics.harvard.edu/faculty/nunn/papers_nunn Nunn, N. and L. Wantchekon (2011) "The Slave Trades and the Origins of Mistrust in Africa," American Economic Review, forthcoming. Olson, J.S. (1996) The Peoples of Africa. An Ethnohistorical Dictionary, Greenwoord Press (Westport, Connecticut & London), 681p.