Radio education and student repetition in Nicaragua

World Bank Reprint Series: Number Ninety-one Dean T. Jamison Radio Education and Student Repetition in Nicaragua Reprinted from The Radio Ma/llicinatics Project: Nicnragua, 1976-1977, Patrick Suppes, Barbara Searle, and Jamesine Friend, eds. (Stanford University, Institute for Mathematical Studies in the Social Sciences, 1978), pp. 199-221  Chapter 7 RADIO EDUCATION AND STUDENT REPETITION IN NICARAGUA DEA N 'F. JAMNISON 7'TIe World Bank IN MOST low-income countrties, studenits Nwlho have repeale(l onie or imiore grades occupy a substantial portion of the availabl e elemeni taryv-school pla;ices. In ad(diLi(,nI to the lpsNc1iolobgical and perhaps ecornomic cost to the chlild that this rel)p lioll enltails, thlere is a penalty to the school systo.um: (lass sizes tend(i to iciicrc;seC, S(10 \oI l)Lclglets rise. and access to anyv schiooling at, all cxtemltts 1iTio slowly to hitherto excitlded Imptilatit ins. It is thuis stm plr ising thtat thle literal11 til contains few stu(ies in(deed f the (l eteIllillants of stlidel ill tihre aml r epetitioni' ancl that proj ect t ValltitOllS tail to rIeport Oil the impiw t ol repetition of the iitervenitions thev aIssess.2 T2he literatture that is The author is indebted to Marian Beard, Jinwesine Frienid, Edward G(eorge. Roger (;rawe, Robert Horinik, Emile k .\nian%. Kathleen M\f N,ls%. Francois O)rivel, Barbar'a Searle, Anita Summers, and Patrick Suppes For helpt10 (tis(-cussions concerning aspects of this clhapter, and to John Sheehan for computer programming. Earlier elrsions lof this paper were presented at a Septenmber 1977 uLouleiet c. IF(lut anil and Economic Development in Latin America." sponsored bv the Program of Joint Sttu(lies of Latin American Economic Integration IE(1 El.) and the Nationial Btureau of Economnic Re- searcl, and at a meeitng of the Comparative and Internationial Education Societv, M\exico Citv, March 1978. TIhe conclusiolns expresse(d here are ntOt ten (sc. il s thlose of the World Bank or the Agencyv f0oi International Development. I use failure to deniote a stotleiit's not l)eing promoted to the next grade anii u1eintlim to denote his both fiailinigt anid I evi id liig, the tbleqnent lil eal. 2 A special issue of the Butlletini of lile International Bureau 0/ Edluration (IBE, I 9W)1 contains an aninotatedl bibliograp.l c concerninig waqon c ill education; IWe term v 'ol 7 tu' includes b)oth dropout and rel)etition in an educational ssste m. Onilv a lfew of the books and( articles that tlhe sp(e ial issue annotate(d (ealt in ants Ns ay with causes of repetitiotn; of the fewstudies (ited, most dealt with European (counitries I Belgimn tn, France, Gel Illanii, Italv, and the UTSSR) and olnly a few witli low incomile cotuntries (India andcl 'I'haliland). WlAlT (1971) provides a more recent study o. odeterminiants of repetition with data 1romi 199  20() JAMISON a\ a-iihible concerlinr repel itioi sulggests, lau)isibly Cenou.gh, that give-nl a school system's promnotion polihc, a chii(i's per foroiiance and atten- dance at school and the economnic (-llditi()ns of Iis baick'gruind plav ilipora limt roles; this Iiltenitii i-c is more stiggesti%"e than definitkiic, more (jtialitaitivc tlhatn (jia titative. Though little is known about the detcrminina uts of re)etition, a good deal of sta.tistical infornmtion is available concerniing its extent (IBE, 1972; UNESCCO, 1974). To give a sense of the magnitude of the problem, Table 1 shows recent rep'.tition figures for a number of Latin American c')untries. The numbers there are high indeed, though typically only half as lhighi for the fourth grade as for the first. It is the magnitude of these numbers, and their imnportance for the economics of school systems, that suggests the importance of in- creased attention to understanding the causes of repetition and the impact of educational interventions on its extent, My purpose in this chapter is to analyze the impact of the Radio Mathematics Project on repetition rates. This analysis provides in- formation not only on the project's effect oni repetition, but also on the other determinaints of repetition and how to aissess themi. To undertake the analvsis, I construct a statistical mo(lel that first p)re(licts posttest performance and then uses posttest performiance with other variibles to predict student fi'ilure, repetition, and reenrollment. The chapter is organized as follows. Section 1 briefly sti ini ariizes previolus evaluations of the Radio Mat'hemiatics Project. Section 2 then de- scribes the model that I use to assess the project and the data used in the model. Section 3 contains the statistical results and Sectioni 4 dis- cusses their economic inmplicatioins. Section 5 pirovides a brief coluhl- sion. 1. PREVIOU-S EVALUATIONS The Radio Mttheimatics Project eVillUtecl student cognitive per- formance at the first-grade level in both the 1975 and the 1976 school years (Nicariigua ii school years run from February through Novemnber). The 1975 ev-aluiation was prelim-iinary: TIhliere wtas no ca undoni assigninient of clissr-oons to radio versuis co ntlol, the radio curriculum was still being developed, the samniple size wals relatively B1 ,iil. I lwe (Y ot il(lie's on detcrnillioiallts ot pt p itiotl co ltrasts itht.111 *o cteltsive litel'atuir'e otl (dtetm'nin.ntn o torlropping out, 00('1 ht ofvhiidi is annotated in the 13BE biblihg ig *'I'' a.l(l a ell sulllniarizedl ill Beir'ni, Kinwsy'V *andI Mt ( ;inin l9)2j. Jallade (19177 .1111d jamisorl l Iand L X, clpll. 8) latv' et)1i(t'(l ediden(e that thle e'ofloillic conmlepueC8e of rep)etitioln are suhstantial: theji Nvatk N; gg!ts thiat I hr internal late of retuirni to investmnenit ill schooling tend, to he highilv sensitive to tie revpetition rate.  STUDENT REPETIT ION 201 TrAB'L,E 1 Repetition Rates in Latin American( Countries Year Grade 1 Grade 4 CF N I R.tI. ANIFRICA \Nt) CARIBBFAN Costa Rica 1972 11 4 C(uba 1971 41 27 Dominican Republic 1969 33 11 El Salvador 1972 16 5 Nicaragua 1971 15 8 Panama 1972 21 10 SmuiI{ A1EFRICA Argentina 1969 23 9 Chile 1972 13 8 Colonibia 1972 22 10 Ecuador 1972 16 10 Guvana 1971 17 7 Paraguav 1968 26 9 PerC 1970 25 12 Uraguay 1972 30 14 Venezuela 1965 24 14 Note. From L'NESCO Statistical Yearbook (1974, pp. 203-204), "The enitries in the table are repetition rates, expressed in percent, where repetiton is calculated by dividing the number of repeaters by the enrollment for the same vear. small, and, for some of the schools, the occasional presence of project personnel for formative evaluation may have affected the results. Nonetheless, reasonably good data were collected, and Searle, Friend, and Suppes (1976, chap. 9) report on an analysis of the effect of exposure to radio on mathematics achievement. They found that students who had had radio lessons scored about 67% correct on .he end-of-year test, while students who had had no radio lessons scored only about 55%; this cliffer-ence is highly statistically significant and reflects a difference of about half a standard deviatioln. The evaluation of the project based on the 1976 data (reported in chap. 3 of this book) provides more conclusive results than do evalua- tions based on the 1975 data, The sample size was larger, the stucdents wvere randomly assigned to treatmnent and control groups, and, pelhaps most imlipolrtant1, a tested and revised version of the cur- riculum was in use. Table 2 (reprintedl from chap. 3) contains some of the more important results. Perhaps because the curriculum had been  Student- ,TiL 11 1 characieristics Age Matriematics Dropped out Maihemalics FNiled Reenrolled Sex pretest Attendance within year potetfrt+March 1976 I F'arents' occupation scr ae(0, 1) Nv:9 (1) (0, 1) I Distance from school March 1975No195( ) Number of previous repeats - School characteristics Exposure to radio Teacner characteristics Class size, etc. FIGURE 1. Causal modiel of radio's effect on learninig, dlropout prIoCellsity. and flailure pwopcy sity.  STIU'DENT REPETITION 203 TABLE 2 Results of 1976 First-gradle Po,ittc(s in \hithemiutic > EXI) I ilIVIIAl olltl ol (radio) Number of classes 30 23 Number of students 571 438 Number of test items 84 84 Meati item score-' 65.6 40.6 Stanidard deviation 19.9 18.9 " Percent correct. improved, the results showv a larger difference in favor of the radio students than was found in the 1975 data. On the posttest, the radio students performed at a level a full 1.26 standard deviations higher than the controls; thev anse ered half agaiin as maniy (ueistioins correct- In this chapter I extend the scope of the ev'al uations just described by examiningi, the effect of the project on students' propensity to repeat or drop out. At the time of writing, data on these outcomes wvere unavailable for the 1976 school year; I therefore use data from 1975, when the effect of the project on achievement test scores was less pronounced. 2. THE MODEL AND DATA The Model Figure 1 diagi-anis the basic structure of the model I use to assess the impact of radio on the various dependent variables of interest. The two large boxes, labeled "ctudent characteristics" and "school characteristics," contain the exogenous variables of the model; the six smaller boxes containi the six endogenous variables arranged as a recursive system. With six endogenous variables there are six regres- sion equationis in the model, and the arrows leading into each box containing an endogenous variable inclicate the independent variables to be used in the equation to predict that en(logelnous var-iable. Note that the recuirsive structure of the model allows one to estimate indi- rect as well a,, direct effects of one variable on another. For example, the diagram iidicates that radio may influence posttest score directly, or indirectly through whatever influence it has on attendanice rate and  204 JAMISON the subsequent direct infl Uencc of attendance rate on posttest. Sihnilar chains of indirect and direct influencea llow estimi ti C...n of the ef'fect of radio on failure and r eeiirollriient. Economists label analysis of the deteriniiaits of t(hievelncilt test scores "eCCILcational piod(lction ftitictioIi" Stitdies; Lau (1978) pro- vides an excellent discussion of the intliods tIse( in (dl1icati(jilid production-functioni analvsis and observes that miiost such a nalvscs. assume a single output. The regressioni indicated in Figuric 1 in which posttest score is the (de1)eplndewi variable is an examiiple of tlradli- tional educatinial pod uction-f inctioii analysis; lhove er, the llmodel as presented in fact has two i ndic:ators of Studllt IperfOrmaf'ICe-his posttest score and whether h.e f;iled first grade--and is thlUs properly considered a multiple-output production model. Viewed fil l a pro- duction perspective, however, the model differs from)ii standarld multiple-output models in that one output, posttest score, is both a final output and an inteirmiediaite outptut because of its effect on propensitv for failure. Three of the dependent drlliales-ms'ithin-vca I ('l)io t, failure, and reenrollment-are dichotomous, that is, they can take on1 onlyv one of twvo possible values (indicated in Fig. I as 0 or 1). Spec(ial statistical tech niquLLc, are required to hancdle regressions with dichotomous clepenilenit variables, ancl before I (liscuss further the variables in the model, I shall describe the regression specifications that will be used. The regressions with cont;lu ouis (dependenit variables-those for pretest and posttest scores andl the one for attendanice-are sp)ecified in the standlar-d way: F = bh