l/i./ V 1 A p "blitot,on of Th« Celltgt of Agriculturt I \ \J U N I V E K $ I TY OF CALIFORNIA '-"hun " The seventh report in a series on Efficiency in Fruit Marketing ECONOMY AND ACCURACY IN ACCOUNTING TO GROWERS FOR FRUIT RECEIVED AT THE PACKING HOUSE B. C. French and R. G. Bressler LIBRARY UNIVERSITY OF CALIFORNIA DAVIS CALIFORNIA AGRICULTURAL EXPERIMENT STATION GIANNINI FOUNDATION OF AGRICULTURAL ECONOMICS Mimeograph No. 1 49 June 1953 1 FOREWORD This report is the seventh of a series aimed at improved efficiency and lowered costs in the local marketing and packing of deciduous fruits. The present report deals with only one aspect of packing costs — the costs of accounting to growers for fruit received at the packing house. Two commonly used methods of accounting — the separate-lot system and the sam- pling system — are considered as to their costs and accuracy. Studies in sample plants indicate that the costs of separate-lot systems vary from $0.13 to $1.09 per 1,000 pounds of total plant volume. Improved procedures could result in about a i*0-per cent reduction in this cost for the average plant. Sampling systems also show wide variations in cost, depending mainly on the desired degree of accuracy and the size and number of growers. The problems and procedures for efficient sample design are discussed, and the costs of sampling systems are compared with the costs of separate-lot procedures for varying conditions. These studies were made cooperatively by the Giannini Foundation of Agricultural Economics, California Agricultural Experiment Station, and the Bureau of Agricultural Economics, U. S. Department of Agriculture. They were made under the authority of the Research and Marketing Act of 19U6. Tgoas.jEUTls tevcaqmi is ••tents zbli^-s 1« 'siihsv^k &di zt 'Hoq&i ' 'eirilT • c i¥* • '^'^ • ' "~ ."ii ■■''jeq ferns 3a 'sJiiaffi 'J jssoX grf^ ni 3&zcy teiow^Z bns siaoa edJ -Ti»ao«! ^til-Aoaq lr> ~&Y9q8s eno ^Ixt6 "n>iw slash" iioqoi ^m»eysq er.T tosq erfif -,ta batf-ioos^ ' J-xxrtl tol aiewsjj oi gai^nr/ooofi Tta" -assa- sff^t -brie rT6.ja^ ioI-9^s'ij;q&'?.- sji^-» , 3ni^rit'Oc ; 'ij3 ,i lo- 'crnrl^em bsW /IxibtTruoo .Yofi.ti/coii -fans aisoo lizrfi- a?' as betsMsiJco 9is--W,tB'Ys gniiq 'Wl-e^ieqs'a It? e.taoo 9ifi J-ari* s*j.fe6xbrii e#a$Iq slqataa at aaxbhJ3 t&HLiLov chimin Jc^T'it'- ^ ?O.J[ft fti ?£ OS nj'Vil; vkv •'•tis^pV? tifU ni rroWorfbeT Jrjao teq-Oii s toodi ni *I«5s*i dliroo 3frr.tffa90c.1q bavcnqntl ertoiJxx'isv ^biw'wivte oe£b- 8tH»ieta sailqfit&S .Jnslq ^•asi****' erii tol Jcot ol-wttftespefq boa ?ir,9ldo'i^ oriT '•arts'vyxg 1« isd-nim hnn ^nxiqffifss c'daoo- &di* bait. , baaatioaxfa sis nsxafch .•rTo.xJ-sJR ..tosmi-sonxS ffe*iitf£i/3. (HE :not£98 >.cb-'is geeriT <*r<3 'lebnir s>bsrt ?T3W' '^sdT Efficiency in Fruit Marketing ECONOMY AND ACCURACY IN ACCOUNTING TO GROWERS FOR FRUIT RECEIVED AT THE PACKING HOUSE B. C, Frenchi/ and R. G. Bressleri/ In the operation of packing and processing plants for many types of fruits and vegetables, field or orchard -run products are received at the plant from a number of growers. To account accurately to the growers for the products received requires determination of the weight or count for each of several important grade and size classifications. Two methods are commonly used to obtain such information: (1) the separate-lot system, where each producer's product is kept separate from all others as it is handled in the plant and (2) the sampling system, where a small portion of each lot is selected and the distribution of the entire lot estimated from the proportions observed in the sample. With the sampling procedure, the fruit of different growers may be comingled in plant operations. The design of such systems has an important influence on the accuracy of accounting to growers for fruit received. In addition, there are im- portant effects on plant operating costs. The first section of this report discusses accuracy and costs for the separate-lot system. Following sections deal with sampling systems and with the comparative plant costs under the two methods of accounting to growers. THE SEPARATE-LOT SYSTEM The essential characteristic of the separate-lot system is that the individual identity of each grower's product must be maintained until it is sorted, graded, sized, and the amount in each size-grade determined. In order to prevent the mixing of products from several growers, each lot is run separately with a delay or break in plant operations at the end of 1/ Cooperative Agent of the University of California Experiment Station and the Bureau of Agricultural Economics, U. S. Department of Agriculture. 2/ Professor of Agricultural Economics and Agricultural Economist in the Experiment Station and Director of the Giannini Foundation. t fSJ>I'VPj*)?s07 D . •■- ♦ swotg Id -'isiiftiin v.-" moil ia&ii\ »b ao^xoHsi l >5vjc909i aJoirfjoiq ©iicf $ efttirtj} onsdnrtqmx Ie-£aV9& "lo do*? :>lni rkws nxBtftifo of fcoair vjfoo/niccb lo 8&q^ 'crtnm' ib'i a>tnslq §nxs3900iq bns saWosq lo ' aos. Jfe^qfc* sriJ- 'nit • tevHoih ' srcB e« oii&6'«q --fiin-fc-isrir tbl OTsWOtg srfcf-bj ^la^a-ix/obs 5 ituJooas c i&x .tniioo 'io ^rijjisvK jrid" 2c •'aoiisnim'ji sis &bod«."*n ofv'f • . anrx^BOl'iiasfifo s3xe f t «S'vfH^e- Joi*$?f5*r6!iee gito (X) :n6i\t&i ai:'.j-x ac 8"isiilo''ll6 moil ovfr/ioqea $qs iibl^ibq 'xisma jV visit* jjsais^ij gniXqaf S^isml jsB'.toC'-o'iitria- yri^ 'lb' ndxJadx««'aj:b 'uflrf'bnrB 'b^lsjX'sa -ai «toX rfos's lo ,*-fi»beoO*iq grs rXqpis'6 odd" riJxW .sXqcwa silJ at -bsfv$JBifc' 8nox44oqcreq srftf «X:ani8Joo- so" '"$sft atswcTEg SnMivilb lo iisrt*t edst" \::>5fi/oo'i» 9iit no' Sonoonni *4^tctep^ ne sert ams'tfavje lo'agzeob erf?- -rfx • ens s-iorfi • t sbZHbbs al .bavlsosi jiw-xl sol stiiroi^'oi prxJiu/coss lo. t-xoqyi sMfr-lo ndlioea %*s ixl srf? ;s,isob jafchwaqo '^riitlq :nc zioo'tid id&taq StfiwoJ.iolT- ,.«oo2\a JoX-.s-tsoaqoa ©dtf so': ^iaoa 'c.-ns "v^^ooos .-foaaibafb ad-ebo '^nsXq ^Vxi^sf^oo'yiiw + f«-xw bits'' Java §rt££^3 Wjx Xssb enafroVe 1ST; Itf XO iX ixJ-n ©dJ- fcr&Jraod i gas on j t. . ... f . . ... • . . , >tq a'Tawoig -rtoss io x jh&oX • A sufcxvibnf v.uoatB bAjt • ban ' t basl'a -< t bdbsig : - t b& J'i03 ef ybvzq lo^arfSatlai silt ^xi^varzq cj 'Tabid ni >l£flTff' icJ s tBl^t b ri5'XK"'YleJ'a*f£Sqsa nih 'ax each lot. During this delay period, the amounts of product in each of the several classifications are determined and recorded. In fresh fruit pack- ing houses, the delay also permits the packing of products remaining in the bins or packing belts so that the actual packcut can be determined. Ideally, such a system should provide a very accurate accounting for each grower subject only to such minor errors as result from weighing procedures and the estimates of the tare weight of containers. But a complete separation of lots would usually require relatively long break or delay periods, and this would greatly increase plant operating costs. In order to increase efficiency and reduce costs, various short-cut pro- cedures are commonly used. These serve to reduce the break-for-lot period, but they also introduce opportunities for error. Most of the possibilities for error stem from the fact that short-cut methods usually do not separate each lot completely, and the amount of product in the overlap is obtained by estimate rather than actual measurement. In addition, errors may result from estimating the weight of fruit in some grades— by applying conventional standard weights to the number of boxes where the standards may differ somewhat from actual weights. In spite of these opportunities for error, practices observed in Cali- fornia fruit packing houses appear to be designed to give reasonable accuracy and to protect the interests and equities of both grower and packing house. Idle Time With Separate-Lot Systems The direct cost of a separate-lot system includes the cost of weighing and tallying the amount of fruit in the several grades, but a more important effect is the impact of the plant delays or breaks on over-all output and efficiency. The interruption of the flow of fruit through the plant means that some workers will be completely idle during this period while others are forced to work at lowered rates of activity. The loss of effective working time depends on such factors as the plant volume per hour, the num- ber of dumping units and sorting-packing lines, the average size of lot for individual growers, and the length of the delay or break-for-lot period. Table 1 summarizes the results of studies of separate-lot systems in a number of plants packing apples and pears or receiving olives for process- ing. The factors most subject to control or change by the plant manager -sat.', fiuiij rt&eixl iflr ibyh-ronsi ckjb bsattnaieb sis eab *Jaoi » ■■"»• - . • j.^\-- to jtfsXvtoaq s/l* siim-ian sals vrslsb ip9!.:i3iUao-.3b e.d ftso JJuorfbsq Z&itioc ad* sad* da aJXeri gniiJ A iw& »p.'iaf:ifi3'noo 3o vdsiet* &&} lo mif fs ' y.ss'td srtoX. ^eyifildnt siinpsn -^LtAueii bXaov trfoX 3o ttoX4 f k^tfOo gAXvt£-i3c^ ^ajsXq & 88810^ yXXr-s-ig bXboi r-bt^ tfyo^r^-Tioijtri auoitsy t 8X8oo soobsi bns t 0f i- •sbijtx'XxdXbaog afb 30 J:abM ;rtoi-» *iq3 eaXXinbiai aisisqas ,*jpjn cb ^LUkpjO aJsartV+sm ii.'o-J-ioda dart; < ba^XadSlp ysX gaXs^Vo aHtf px tfauhriq lo Jiritt&tts i ■ «r*93fb vara ah" dibb.4 dl at. drifted 0. po^iibBiq &tB d$b:f eaJ •s ua& seesaw XsirdoB OfU fti ixi"i3 lo drigisv <*£cd 36 rtadfe&iin add ♦ e-ldaxeti 3 esxixv'L'iioaao grtXaosq •sairfgi;^!* 3o^aoa .sdJ ?.abuX:n:x iijajgfl fn^tio.qmx eican 3 yu6 .. v 8dfjsTj Xs/iavsE ' bns\ tuqiuo Xlo-agvo no. 3vii:.qid 10 -• enGs]u .Irit.Iq .gri} rtsifOT.iJ' J Xx"!** 3c- woX3 eriJ" *>6 no/Vini« ct^tUn plxrfW. bpist? ciri^ snxifb slbx. y,X3.J&'Xqmco 'lof tX fcvX.*n^3 , »9 3d saeX sriT .vixyxdna 3o ee^ei bsnWoX' d ^9X-9vt■s*IBo;98 a 3o rfaoo >lo^iXfe Sill Ut ni d±xw3'3o Jai;oihb 'ttij anx^tXRX bfff, >X'3b inelq h&S 3o ctsacmx giii *?X ••doal'lst .♦nia».^„,-ti;crt isq a««fXoy_ S**iq srfj a » rioi.e no ehit^h est* sniiliov io3. inl, 3o ^x?. 9ss7*?ts 9«> t fe^nxl sriX^Deq-snx^'toa ftrrar sdhix* j>n r qpxfb lb'' led . .box.730 j;o.J-.^G^- : \'r.^«i<3 'ic Ttfiidb 91W 3o rfi^rhX &dj bxiii , uievoi .- 'inubivibni ■•* xtx .^tteiana . .fpXiui^Tteqaa 3.o gQi&uJz 3o atXosei alii "fee'S ir»8ffl8«tt» ,s l 'sfd^T - •M»odtq io3.«9riio saxyiisos? iH atseq Wb ^esXqqe gnfSbiq" fitOlq '3o ; Ttsdiman ;- t»g«irTflT;.c-n8iq .add -^J _5&iti.!o" '10 Xoiirtoo or' ia'aj-due tfaorc B^J'or-fVdT '' ,'ndl 3 TABLE 1 The Effects of the Separate-Lot System on Effective Working Time in California Apple, Pear, and Olive Plants Volume of fruil Lots Break Loss of effective Per Per Per run per time work in 5 time PI ant plant-hour line -hour lot line-hour per lot Per hour Per cent 1, 300 pounds minutes Ai sole and pear packing houses A 20.U 10.2 10.2 1.0 3.0 3.0 5.0 B 18.0 9.0 >Jl 5.6 L 19.8 9.9 a. 3 2.3 3.7 8.6 1U.U M 2U.8 2U.8 19.8 1.3 3.1 luO 6.7 N 38.U 12.8 8.5 1.5 11.3 16.9 28.1 R U5.o 15.0 tu7 7.8 S U6.0 15.3 8.1 1.9 0.9 1.8 3.0 T 31.2 31.2 h.2 7.0 U 60.6 30.3 23.3 1.3 2.8 3.6 6.0 o: ivfi plants I 3.1 3.1 2.0 1.5 2.9 U.5 7.5 II 13.3 13.3 U.6 2.9 2.7 7.9 13.2 III 10.0 10.0 1.8 5.U 2.0 10.9 18.2 IV 11.8 11.8 2.8 U.2 3.U Hi. 2 23.7 V lf)& 10.5 5.3 2.0 2.3 U.6 7.7 VI 7.2 7.2 2.9 2.5 U.2 10.6 17.7 VIII 5.2 5.2 2.8 1.9 U.9 9.1 15.2 IX 12.3 12.3 3.2 3.9 3.6 lU.O 23.2 Dashes indicate data not available. 1 '■ - ■ r — — — j ■-- - ' \ ■ * > ' - j ' ; ■-€.2 • i 9m rT srujiif r are the average size of lot and the length of the break period. In apple and pear plants, the average break-for-lot period ranged from less than one minute in Plant S to more than eleven minutes in Plant N. Individual break-for-lot periods ranged even more widely — from less than one-half minute to more than twenty minutes. In the olive plants studied, the break- for-lots averaged from two to five minutes, while breaks for individual lots ranged from one to ten minutes. In both the olive and the apple and pear plants, the average break time was about three minutes. Size of lot is determined primarily by the particular size and pro- duction characteristics of the individual grower. It can be controlled somewhat by plant managers, however, in several ways: (1) scheduling picking and hauling operations from particular orchards so as to result in larger average deliveries at the plant; (2) consolidating several truck- loads from a single grower into a single lot at the plant; and (3) dis- couraging very small growers or by combining the fruit from several small growers into a single lot for plant accounting purposes. The table indicates that average volume per lot ranged from U,000 to 23,000 pounds for apple and pear plants. Individual lots handled at these plants during the period of study ranged from less than 1,000 to more than 70,000 pounds. Considering all plants, the average size of lot for apples and pears was about 10,000 pounds. Lots were much smaller in olive plants, averaging only about 3,000 pounds. Individual plants had averages from 2,000 to more than 5,000 pounds during the periods studied. These averages were greatly influenced by a few relatively large cases, however, and many lots were of very small size. Individual lots observed during the study ranged from U00 to 26,000 pounds; 25 per cent of all lots were less than 1,000 pounds and 32 per cent fell between 1,000 and 2,000 pounds. The table also indicates the effects of the separate-lot system on effective working time. As mentioned above, these effects depend on such factors as the plant volume per hour as well as on average size of lot and the average length of the break-for-lot period, In apple and pear plants, the loss of time ranged from 3 per cent in Plant S to 28 per cent in Plant N. This extreme variation is due primarily to the difference in the length of the break period— less than one minute for Plant S and more than eleven minutes for Plant N. The lU-per cent loss in Plant L, on the other hand, is primarily the result of unusually low 'COOC .v con l «i to! ■ j • •; • i as iifon isq swi/xov J'flBlq erf; v'fM eti't ;&rlifelq €s"3q't> *ro£j O: 5. average lot size and the consequent increased number of break periods per hour. Plant U has a break time per lot that is about average, but with lots averaging more than 23,000 pounds, the time loss is only 6 per cent. With relatively smaller lots and typically more break-f or-lot periods per hour, the loss in effective working time is typically greater in olive plants than in apple and pear plants. The table indicates that the time loss ranged from 7 per cent in Plant I to 2h per cent in Plant IV. The large losses in Plants IV and IX are due to small lot size relative to plant volume and so to relatively large numbers of breaks per hour. Plant I has the lowest average lot size of all the plants studied, yet, has the lowest time loss. This is due to the very low plant volume per hour and the re- lated small number of lots run per hour. The effects of size of lot and break time on effective working time are indicated in Figures 1 and 2 for more or less typical plants. The diagrams are based on typical values for plant volume, lot size, and average break time. In individual plants these factors may differ considerably from typical values. In such cases the time loss would differ from that indi- cated by the diagrams. With these limitations in mind, the first diagram (Figure 1) indicates how increases in the average break-f or-lot period cause corresponding in- creases in the time loss. In general, each increase of one minute in break time is associated with an increase of 2.5 per cent in the loss of effective working time in apple and pear plants. Because of typically smaller lot sizes, the effect of increased break tjme is more pronounced in olive plants— each one-minute increase typically results in a 5-per cent loss in effective work time. Figure 2 shows the general effects of changes in average lot size on effective working time. Since increases in lot size usually mean fewer lots per hour and so less delay time, this diagram indicates that the loss of effective working time becomes smaller and smaller as lot size is in- creased. In apple and pear plants, lots as small as 2,000 pounds would mean an average loss of nearly ho per cent in effective working time. With lots of 10,000 pounds (approximately average), the time loss would be re- duced to 7.5 per cent while larger lots would bring further reductions, approaching a minimum time loss of 3 per cent. In olive plants, lot sizes of less than 1,000 pounds—and 25 per cent of actual lots fell in this category —would mean time losses of well over h0 per cent in typical plants. 0 4 8 12 16 20 24 Size of lot in thousand pounds Figure 2. Effect of size of lot on the utilization of plant time. 7. Even average lots of 3,000 pounds are associated with an average loss of 15 per cent. Very large lots can be handled quite efficiently, however, with average time losses approaching 2 per cent. Costs of the Separate-Lot System As explained above, the major impact of the separate-lot system on plant operating costs is in terms of the loss of effective working time and so in the reduced volume of fruit handled per hour by the plant. In most plants, the elimination of the separate-lot system would permit only minor changes in the working forces — grower tally girls for packed fruit could be eliminated in fresh packing houses, and the number of men weighing and handling graded and sized olives could be reduced in some olive plants. Table 2 summarizes data on plant volumes and estimated direct labor costs for the apple, pear, and olive plants included in the study. This table shows that the elimination of the separate-lot system would result in increases in the potential plant volume per hour, with small increases where the present system results in small reductions in effective working time and large volume increases where present time losses are large. Most of the plants would be able to reduce the direct labor payroll per hour, although these changes would be relatively minor. The combined influence of direct labor reductions and increased volume per hour would be reduc- tions in average direct labor costs (exclusive of packing labor and other piece-rate workers) ranging from $0.13 to $1.09 per thousand pounds of apples and pears and from $0.30 to $0.80 per thousand pounds of olives. While these costs of the separate-lot system may not seem large, they become significant in terms of the total volume of fruit handled by a plant in any season. Moreover, the range in costs emphasizes that many plants can improve efficiency and reduce costs by adjusting their separate- lot systems in order to minimize the loss in effective working time. These adjustments may include the installation of automatic counting and weighing devices, better coordination of tallying activities in the various parts of the plant, and improved systems of record keeping. The potential costs of the separate-lot systems in the sample plants with the break period between lots reduced to an average of two minutes are shown in the last column of Table 2, No improvement is indicated for Plants S and III since the break time per lot is already two minutes or less. v;lnc laJ&YS^oX-***? sq*»n ad* lo .fcj&qpd: ^t«« sffJ.«#vtfcte bsmiiiXqxa a f gfiiAiov sv-fcj03lls» lo BaoX arii 1o «mT^ rtx ai a$aov snionaqc .iJnalq anJ " , £y-ii/oft .?aq; biiXbajd Jiiral $q afl&Xdv baoubaT aitt nx irn^q J^Lcrovrmefa^B ■tbl-siaizqise eritf 5o «oJ#snJ/axXa adi • # .atnoi u bajtoaq iol eiTcxs tsvcxg— a^a'iOi anitfiov sriJ* ni as^nisdo r itaic 1o ^dsiua aiW Jans , a sauori gnijioeq ria-dil at ba&'x.inXIs $d j'.'tl.o sraoa (ix b&oibai -ori blyop •.*»*•*> iio b^sxe f»ns babsijj gniXcfl • todeX Sosy'ib boismxige bns e&mulov j:t=.Icj .no -ndrrb s-jsiismnt's S sXrfe aid'f . .\;b5fla.ei{j -ni fejbwloni atasXq svxXo bns v iBoq t aXqqs erft ^ol &£uq&i .bXuow. waJ.wte xol-a^suqde «dj -:aoxd saimxXa sxiJ rfcx ■saefi^ionj" XXsrae alxw t wod i3q -jhuXov tfnisXu' XjsxtfnaJ'oq arii ni .^nik-sow sv±JOi'Mj:-.ax snobiojAax XXsma ni aiXxra^t m^e^s- tfn©« #8oM . . cjgisX sis" aaaaoX emii -Jna&wiq oisriw eaanciwix jjswlov s Viaort* 19 q XXoi\.5q TtodcX iamcb sri^ aocb-ji 5.t dXde -*d jrdaov . aoR'.»uXini baaxdsrtoo -idT Ionian tvXavi^al'd'i.-Bd XXxjOW asvisito s .U-C3 h avioda a Ids J eear-sionX nx c«*vs .nx aneii ■^.o ajjfijoq 1 . . ■ : ■ 1o -jm !4.',0«*ibo v i1 bni 3«x*idw .- •• . ; • arfT .'.gnJ OWcda C-1I to noxJsni cicoo .••jeJvted 4 s$oxvbb -gnxiis JS^e bavoi;{itrx bns^nsXq ad-j to eJicq ismsvoiftax oU .'.2 aidsf .?io. amnXco .tesX Ds»«Xi"/ ai jaX Tsq-srax. i MeBld s>At -aonie. w fena inaXoa 8 TABLE 2 The Effects of the Separate-Lot System on Potential Plant Volume and on Direct Labor Costs, California Apple, Pear, and Olive Plants Present — with separate lots PotentiaJ . — no separate lots Labor cost per 1,000 Plant vo_Lume Labor Plant volume Direct T.a Vinv XJGL D Ul pounds for the separate-lot system Plant 1,000 pounds per nour plant payroll per nour cost per 1,000 pounds 1,000 pounds per hour plant payroll per hour cost per 1,000 Present Improved : two- minute breaks- for-lots Apple and pear packing houses A 20.U $ 62. U0 $3.06 21.5 $ 61. U0 $2.85 $0.21 $0.15 B 18.0 73.90 U.11 19.1 7-2.80 3.81 0.30 L 19.8 82.80 U.18 23.1 81.70 3.5U 0.6U 0.35 M 2U.8 58.70 2.37 26.6 57.70 2.13 0.20 0.13 N 38 .U 1U0.00 3.65 53. U 136.80 2.56 1.09 0.20 R U5.0 119.70 2.66 U8.8 117.60 2.U1 0.25 S U6.0 13U.20 2.92 U7.U 132.10 2.79 0.13 0.13 T 31.2 103.90 3.33 33.5 102.80 3.07 0.26 U 60.6 120. Uo 1.99 6U.5 117.30 1.82 0.17 0.13 Olive plants I 3.1 11 .Uo 3.69 3.U 11.U0 3.37 0.32 0.19 II 13.3 2U.90 1.87 15.3 23.60 1.55 0.32 0.25 III 10.0 28.70 2.87 12.2 25.30 2.07 0.80 0.80 IV 11.8 21.00 1.78 15.5 19.10 1.23 o.55 0.35 V 10.5 25.20 2.U0 11. u 2U.00 2.10 0.30 0.28 VI 7.2 17.80 2.50 8.7 16.60 1.91 0.57 0.32 VIII 5.2 17 .Uo 3.35 6.1 16.30 2.67 0.68 0.38 IX 12.3 21.30 1.73 16.0 18.80 1.18 0.55 0.35 a/Dashes indicate data not available. 1*0 : 1 . 11. Sample Size The second important aspect of sample grading, and the one of primary concern in this report, is the size of the sample to be selected. With any sampling system, the estimates of the proportions of the fruit in each grade will be subject to some error because the distribution of grades in the small part of the fruit that is sampled may differ somewhat from the true distribution of grades for all the fruit delivered by the grower. The magnitude of such differences can be reduced, however, by increasing the size of the sample. The objective in designing a sample grading system is to provide a basis for making accurate payments to growers for their fruit — to limit the possible error in these payments to some specified amount with a small probability of exceeding this limit. The practical sampling procedure to follow depends on the amount of advance information available concerning the total season or pool period deliveries for each grower, the expected proportions of fruit falling in particular grades, and the price for each grade. As knowledge about these conditions increases, sampling systems can be designed that are more efficient in achieving any desired degree of accuracy. Four cases regarding the amount of advance information may be distin- guished: Case I — nothing is known about any of the indicated conditions; Case II — reasonable estimates can be made concerning the expected minimum total quantity of fruit to be delivered by each grower during the season; Case III — advance estimates are available concerning the proportions of fruit falling in particular grade classifications as well as total deliver- ies; and Case IV — estimates are available concerning the prices of each grade of fruit as well as the proportions and total deliveries. Under Cases I, II, and III, it is necessary to express accuracy in terms of the weight or proportion of fruit in a particular grade rather than in terms of payments. In Case IV, however, accuracy may be expressed directly in terms of total payments or average prices. Accuracy in Estimating the Proportion of Fruit in Each Grade . — Growers and plant operators will ordinarily be interested in obtaining some degree of accuracy relative to the total quantity of fruit delivered by the grower in a season rather than for a particular lot of fruit. Under the conditions of Case I, however, the best that can be done is to sample so as to obtain ^raotiq lo t>f»c ■ ??i 3ni»"':-T; %£qr-se to >oa-qa>; ctnact'ioqa'x bnoose ©*E ■i'i'i'iV vjbdiPtelda 3'iri? no iJ'i-rifro-lru yonsvfce baooyqxs art'* •iie&woia nos& lot aaxtfovllerb rfoB© toi sr&'^q arlst bfla taafxarajj itXiroxiic^ lo &9^3^b bsixaab -yjab jjaivdxrios -At 'n&Xoxl : gnxti^x ; anoiJibrtCo' bsJcoxfin uotfa ffwc B±5C! 9!W rams iiriy no eonnqsn woXXo'r I Xooq to ft0BB0B Xsiov a'riSr :IXsl Jxinl lo artoitf'ioqc'iq tj-'ods sgbaXwonn ;>A ts'bjfca li §ft£fftcs$9$ eaa tfanxisi aXdanca - ■^svMsb Xs^Ov + as XX3W as :ffrJ69 "Jo s&'jiiq rfrfi 30. ■■xsbnU •.^-jxisvxXab Xj t'ftt 1c aflft^J nx ^SsTtroo! am fed nx oaiW '•tsri&e* gj sXiteirfaaq a fix * irtiXUx txi'x'i lo absia lo iiv?mvr-q o i aerf* 12. a given amount of accuracy for each lot that is delivered realizing that, if the grower does deliver more fruit, and if additional samples are taken, the over-all accuracy of estimate will be somewhat better than indicated for the particular lot. On the other hand, if past experience plus the personal knowledge of the plant manager and fieldmen permit a reasonably close estimate of minimum deliveries for each grower (Case II), the sam- pling procedure can be much more efficient. For either Case I or II, the quantity of fruit to be sampled depends on (1) the desired degree of accuracy and (2) the quantity of fruit in the lot or the expected total quantity of fruit that the producer will deliver during the season. The accompanying diagram (Figure 3), which may be used for both Cases I and II, indicates how the percentage of fruit to be sampled from each lot changes with the size of lot or total deliveries per season for several designated degrees of accuracy. For Case II, a total sample of the required size could be obtained by taking the entire quantity of one or two of the lots of fruit delivered by a grower as a sample. For adequate representation, however, the sample should be spread throughout all of the grower's lots. Therefore, the total sample is obtained by sampling a percentage of each lot of fruit delivered. It is assumed that each lot sample is a random representation of the total of the fruit in the lot. Accuracy is expressed as a per cent of the total weight or number of fruit of all grades. For example, if the admissible error is to be limited to 1 per cent and the true proportion for the particular grade to which the error limit refers is 50 per cent, the estimated proportion for this grade based on the sample may be expected to fall within k9 and 51 per cent ($0 t 1) in the great majority of the samples. The basic calculations are made in terms of numbers of individual fruit and are generally applicable to most fruits. Since the quantity of fruit is usually measured in terms of weight, an appropriate ratio must be used to convert to numbers. Gravenstein apples and Bartlett pears for example, average about three fruit per pound of field run produce, while Sevillano olives average from forty to sixty fruit per pound. The lower of the two olive figures should be used in estimating the number of fruit delivered by a grower to be more certain of limiting the error of the desired range. Figure 3. Effect of the desired degree of accuracy and total quantity of fruit per lot or per season on the percent to be sampled from each lot for conditions specified under sampling Cases I and II. To illustrate the use of the diagram, suppose that under Case I con- ditions in which no advance estimates are available, a grower delivers a single lot of fruit consisting of 10,000 pounds of Gravenstein apples or Bartlett pears (30,000 fruit). To be reasonably certain that the error in estimating the proportion of fruit in any grade will be less than 1 per cent of the total weight of fruit in all grades would require a sample of 2U.3 per cent of the lot (see point A in the diagram). On the other hand, if minimum total season deliveries for this grower can be estimated in ad- vance (Case II), the per cent to be sampled from each lot can be greatly reduced while still obtaining the desired over-all degree of accuracy. For example, if expected total deliveries were at least 30,000 pounds of apples or pears (90,000 fruit), 1-per cent accuracy could be obtained with a sample of 9.6 per cent of each lot as indicated by point B. For growers with deliveries of at least 100,000 pounds (300,000 fruit), the same accuracy could be obtained with a sample of only 3.1 per cent (see point C). For growers with deliveries above U 50, 000 fruit, use the scale in the upper right-hand corner of the diagram. In addition, it is clear that, for any size of delivery, as the limit of admissible error is increased (accuracy reduced), the per cent to be samples from each lot decreases. For example, to limit the admissible error to one-half of 1 per cent for a grower delivering 300,000 fruit would require a sample of 11. h per cent of each lot, but to limit the error to only 1 per cent would require a sample of only 3.1 per cent of each lot. It should be emphasized that the limits of admissible error shown in Figure 3 do not represent the amount of error that will occur in each esti- mate but rather upper limits to the error that will be reached or exceeded only infrequently. The limits of admissible error in Figure 3 are based on the probability of obtaining an error of estimate as large as, or larger than, the amount specified only once out of twenty times. This ratio, though arbitrarily selected, is commonly used in statistical analysis. The great bulk of estimates would be subject to errors considerably less than the limits stated. With an error limit of 1 per cent, for example, the average error would be about four-tenths of 1 per cent. Whatever the limit of admissible error, the actual error can be expected to be approxi- mately one-half that amount in about two-thirds of the cases. Moreover, the error for any grade having a true proportion other than $0 per cent would be somewhat below the amounts discussed above. Thus, these curves iidnoqoi'qgdJ gai^BftiijeB tgiaw Xniod odd lo fh&a o£ add ggftj iaq £ 4 *iS -noo I ©e^^. isfaqi/ s*eoqg4;a. ^mfii^rijco gdJvlo-aw an* stenttuili s.^.BisyiXab *X3woi5 ^dfiJL Ja?y« 9.io. B^aiai^as.^n^vfce -6ft rioidw oi ..ic eaiq^e niiJsna'/BiO lo- ebptfoq OOO^jOI lo Spii^eie/KpO ixtnl 4o Jo ni.xoiis adct Jsdd ni$-?iao yXdsnugxtei 9d ol . i$q ; X oaad assX 9d XXiw 9hst§ xpa xii Sitn'i ' .lo aiqmga a aiii/pei .bluow .sebaia XX a ox i'tMA ,baad. i.9ddo srU aO ,* (fliei^axb arid oi A ,+niftq ». -hs.ni bdJaoX'ies gd a&c lovoig sxrid lol eCc^sv TtfcJasig ad nsj) ioX dosa rcoil bsXqnsa 3d od *j .vjsiirDSs .lo .93-x§f>b Xi.^-tsvo foixB&b 9"di 51 7c .sbru/oq OOO t 0£ d*6»X da 9isw B9xi9Vi.f.ab Xc rid.Xw banieido id bXtfoo x-»*x«t>»s dnao igq-X xoi <$ temzip bsi-mm 8S dox dos* 9«se erid t (iiirti ooo,, cot) g&a>»cj OM* oox (0 doioq 99?) imo iaq X*£ v^lnO lo aXqnae. a ri#Xi .laqqu add ft* 3X60 a arid sew ^dix/i" aXqrcse a rtdxw bgftisddb sd bXuoo ^oaiuoos >00^034 ovod« asxiaviXab dStv e'xsvoigi i-eS .fltBijgpxb erid lo ismoo bnsd-drigii I *!Oi r d£tt? isaXc 8.1 dx t ooxdxbbs ill Beft&*»a£ ax 10119 9Xd-xa«xmfca lo sgp.ssiogb JcX ricsa moil esXqnup. fro 9 iaq X lo tXsri*9fl6 od tons rpTBS S ;>-Uiip9l yd o^ jnsq 53q 9J*J / (beoybsl ^osiifijoe sXdiaaimbe siiJ ^intiX o^ to'i bX-pw J-Xi.rcl 000 t 00C ^piisvxXsb qawo-tg 9 « j-7 ion9 sxU JimiX oj Jod ,d-oX doss lo iast) isq xi.XX lo woX dos9 lo JYI9Q I3q I»P ^LtO 1 lo sXqidse a siii'psi bXiioW J«90 isq X ijXfln ac nwod-3 ions sldxaaiobs lo a^imtX 3di &Bci-i bsj.isedqms 9d HS.ifOrlz il -j do 96 ni mo.oo XXiv JsriJ 10119 ?ai/oi«fi sd* Jflsseiqe'i fort ±*ai aidT. .sajjti* Y,i"n9wcf lo ^xioqoiq suli ■ ti fJtiiv^d-.gbsif \;h5-- iol lci^sSdi eavivo. 9294^ ,,Sifri'X »avod«- bde^ttoaib ^awomB 9d.t*wo£9o^"iadw9MoB ad-'biXfov' 15 do not indicate the typical amount of error in the estimates of the pro- portions of various grades but upper limits to tbe expected error that would be exceeded only in relatively rare cases. If none of the grades approach the proportion of £0 per cent, the procedures outlined for Cases I and II will involve larger samples than actually required to attain the specified degree of accuracy. If there is no way to foretell the approximate values of the proportions in the class or grade containing the largest quantity of fruit, however, these procedures will remain the best available guides.^/ On the other hand, if, from past experience and observations of fieldmen, the plant manager can make a reasonable forecast of the proportion in the largest grade for a given grower and of the approximate amount of fruit that the grower will deliver in the season (Case III), more precise and economical procedures are possible. For Case III conditions, the per cent to be sampled from each lot, as determined in Figure 3 (Cases I and II), may be modified according to the following general rules: (1) If the expected per cent of fruit in the largest grade or class is between 31 and 69 per cent, use Figure 3 without modification. (2) If the expected per cent of fruit in the largest grade or class is between 70 and 79 per cent, or 21 and 30 per cent, multiply the per cent to be sampled as determined in Figure 3 by: 0.9 if the expected deliveries are less than 90,000 fruit per season; 0.85 if the expected deliveries are 90,000 or more fruit per season. (3) If the expected per cent of fruit in the largest grade or class h/ Proportions should be calculated in terms of numbers of individual fruit . If the average size of fruit is the same for all grades, proportions of weight will be the same as proportions of individual fruit, but if fruit size varies among classes, the weight and numbers proportions will differ. For example, a grade containing very large fruit might contain a high pro- portion of the total quantity of fruit in terms of weight, but a small pro- portion in terms of numbers. 5/ These rules are simplified for practical application. See the Appendix for more accurate expressions. 6/ Use minimum proportions if the expected proportion in the largest grade is greater than 50 per centj maximum expected proportions if less than 50 per cent. For example, the expected proportion in the largest grade might be at least 70 per cent or not more than kO per cent. "CIO 1^ £ v'lUU.iQO'^ 'J J.£ vXfiiqnOOS nf3£ S^SX-'yTq 0*1Ofn '^£80} no ! ca&xo .it} 3c , s'75> ^Bsg'isX Sri j rri -tti.fi i "xo sfmai) "vsq fcacfO'jqjcs ari»t aj? -i©q_ artt viqWli/iu ^droo -tgq G£ hnii IS i© ^^aao iaq ?V bfts °Y sesiq io ebaiij iea3*tcX eifa nx JJUftt lo Jfisc i&q bsJo-qxa .»rii nisi loxctrx iUXX.'-m 16. is 80 per cent and above, or 20 per cent and below, multiply the per cent to be sampled as determined in Figure 3 by: 0.8 if the expected deliveries are less than 90,000 fruit per season; 0.65 if the expected deliveries are 90,000 or more fruit per season. As indicated previously, the real objective in designing a sample grading system is to provide estimates that will limit the possible error in final payments to growers to some specified amount. While the procedures outlined for Cases I, II, and III achieve this result indirectly, the mag- nitude of the limit of payment error is influenced by prices. If advance information is available concerning the prices of the various grades of fruit (Case IV), the sampling system may be designed directly in terms of some specified degree of accuracy in the final payments. The calculation of sample size for this case, however, while not difficult, is too complex to present in simple form. Therefore, the procedures outlined for Cases I, II, and III will probably prove most useful for the majority of packing house operators.!/ Sampling Cost The costs of sample grading include the costs for workers who collect, sort, weigh, tally, and transport the fruit that is sampled and the costs of sample grading equipment ..§/ The sampling equipment consists of items such as grading tables for apples and pears, sizing-sorting equipment for olives, and scales for weighing the fruit. Mechanical sampling devices for sample selection, used in some plants, are not included here. Estimates of these costs, based on studies of packing and processing plant operations, are summarized in Table 3. Two levels of equipment cost are indicated for apple and pear plants. For small plants — volume below 20,000 pounds per hour — a small grading table with a capacity for one or two sorters would probably be adequate. Larger plants, or plants with a large volume of fruit sampled, may require a larger table with three or four sorter capacity and with power-driven belts. The equipment for olives would be about the same for plants of nearly all sizes. 7/ A simplified mathematical and graphical treatment of the sampling problem for Case IV is given in Appendix A. 8/ Since the fruit that is sample graded does not have to be run over the main graders, only part of the sample sorting labor has been treated as a sampling cost. Jrtda isq add ^IqisLum ^wolad fans oxia? i&q OS io t aved6 bna in90 I9q 66 ax :vd £ 2 r xLtgi r i oi bsnxiiiisj^b as" bsIqmjSB bd od ;no?.3S8 -ismic'':.nx ban^xaab ad y.sm mads^?; gnXXqibs ari9 anXbivxs oXqjrfaa '-lo 10I Jnoaqxwpa Sfii^oa-gaxsxB t aisaq bne aaXqqs 1tol aeXda^ sxixbsig' Va rtdira 3&Qtvt,b sinxXqiasg XspxnsrioaM .d-Xyi'l *>d>t aixtxlsisw •io , i aaXsbe bns 9X&a t;Xqmfea iol . ^nxea&ooiq bns ^nxjfosq lo eoibuia no >.»8 5 & t aJaoa icaH^'So 39iaitfii83" • a-eoa, Jn^fliqxup-j 10 aX^voi cwT aldsT ai b9sxia/'-ar!ix3 sir .srioxjcieqo 4nsXq W0X9X scsajXcv— pjnsXq XXeflta 10I .arfasXq isoq bus 9Xqqfe 10I bsisoitnX aia • 10 arxo lo'i -^iosqao £5 d^Xw eXcf-jcr snxb^iy XXsms a-liyori ^taq 'ebnifoq 000 t 0£ -. a riixw ?j-neXq 10 ^sineXq lagisJ , >JGi/pabs yi ^tdaifblq bXiioi-r eiaiic^ owi io-:«aiii^ rltiw gXdsj lagisX a 9ii.x r p9i ygm. ,bsXqrasa jxuil lo smiLov dg-isX .•89vXX-o 10% ixmq,tupo, ariT .e*Xad itevxib-iawoq dixw bus yixosqijo 13^103 iwol .Basis XXa yLinoti lo sdnaXq 10I emo& eni iuo^fi ©cf hXifc^: , gnxXqrase sdd- 1o^«afltf«Mt^ Xa0xrlqsi§ baa XaoxtsrasdJ-afli b9xlxXqm23" A "\T ' . A xxhneqqA nX ns>vx§ eX VI 95a* lol^nsXdoiq rfW, !)ta,y.o ntrr ad ©J avaxl ion asob b9bai§ rxqu'ioR &x da'ri.t iXi'il e& ^dnisr \6 ...fi •efi: bai39ld' nead tisd "lOdal •9Ili*Tia ■■if'(Tflt»P Prt'.t Vo .+fR.-r vfrtr. . fi^Tt. ni;* 17. TABLE 3 Estimated Labor and Equipment Costs for Sample Grading m. Apple and pear plants Volume less than 20,000 pounds per hour Volume more than 20,000 pounds per hour Olive plants Estimated installed cost of sample grading equipment, 1952a/ Grading tableb/ $ 65o $ 700 Bench scales — dial type 560 $3Io 560 $1,110 U80 $1,180 Estimated annual cost of equipmentc/ $ 70 $ 160 $ 160 Typical labor hours per 1,000 pounds of fruit in samples 1.9 20 Typical labor cost per 1,000 pounds of fruit in samplesd/ $ 2.28 $ 2.28 $ 2h a/ Does not include costs of mechanical devices for sample selection. b/ Table for small apple or pear plants has capacity for two sorters. Table for large plants has powered belts and capacity for four sample sorters. The olive sizing-sorting table has a capacity of about 150 pounds per hour. c/ Based on a standardized set of annual charges for depreciation, repairs, insur- ance, interest, and taxes; grading tables, lh.7 per cent; scales, 10 per cent plus $7.50 per year for reoairs and maintenance. d/ Based on typical average wage of $1.20 per hour for sample grading labor. 18. The costs given in Table 3 represent only average annual costs of equipment and the labor cost per 1 J 000 pounds of fruit in the samples. Total costs of sample grading depend on the quantity or per cent of fruit received that is sample graded which, in turn, depends on the desired de- gree of accuracy, the total volume handled per season in the packing house, and the distribution of total deliveries among the individual growers. These factors vary from plant to plant so any simple statement of sampling costs is impossible. However, the general nature of these costs and their relation to sampling accuracy may be illustrated with reference to the conditions in a specific plant. Costs for a typical apple or pear packing house and a typical olive processing plant are given in Table h* In these examples it is assumed that estimates are available concerning expected total deliveries per grower but not concerning expected proportions in the various grades — i.e., Case II conditions exist. The percentages to be sampled are thus read directly from Figure 3« The size and number of growers are given in columns 1 and 2 of the table where, for simplicity, they have been grouped in several convenient sizes. This is a practical sampling procedure that might be followed in most packing houses. The per cent to be sampled from each lot for each size group was determined from Figure 3 for several levels of accuracy. These figures have been applied to the total quantities of fruit to be sampled for each size group. The columns are then added to indicate the total quantities of fruit that would have to be sampled by this plant in order to obtain various degrees of accuracy. These calculations, which a manager may make for his own plant, indicate a rapidly increasing quantity of fruit sampled — and, thus, an increasing sampling cost — with increases in accuracy, especially for the higher levels of accuracy. Note that for any level of accuracy a much higher per cent of fruit is sampled for small growers than for those with large total deliveries. In the illustrative pear plant, for example, to limit the admissible error to 1 per cent for all growers requires that 9,6 per cent of each lot be sampled for small producers and only eight-tenths of 1 per cent for the largest producers. On the average, 2.6 per cent of the fruit would he sample graded. While a uniform sample of this size would give the same total sampling cost and approximately the same degree of accuracy (1 per cent) for the average-size grower, the estimates will tend to be less accurate than the desired amount for small producers and relatively more lo cfeoa £ surma S3S»»*& 3tiMto% 116 yjfl'*^»b t 4riU* ft£ ^rtViiiw fcefbS-c^ &IqrW&& J Art* bavx ids't t &3iM gdWSbq *d» ett «os*9« tad; feeibaSli »taL«Cdv Is*oJ fdtf ^isxiiucse lo asia •siaWo'i? Xatfhc'Ufcra ^Wxse editatifdb £ej©4 lo aaxj^di-tfaib ©rtf bn» sailcpt&e Id tfdsflwlatfa aXq&fcs \aa an * Istavaa /ii b&qj'afg nea6' aVsd Ijed* .YiioiXqrtX? icl >eiad* s£c.>w6X/o*> a3 fcfeltt jfsd? ^■xi/fcfo6*fq giiiXqeu-a XcaWtfeiq a ei exdT .a^sia fasa •' " : i < •/Sat/iitf*? 16 elavsl XotaV^a £ •.iiogil moil banXzrmtfar. asw 4W613 aaie ad* i-^gaifcfti o* bibfte deri* iws Oira'td.oa ddT .qwoig av?£a dosa id! bolqttisz at Snkl? giii* yd balqmsD ed &t ovsri blx/ow cfsd^ ^xwil lo a-^i.Jnsitf Xs^oJ £ ridirfii t anoxi eXj'sI-l-o aaadT .ypniyoos lo aa&i*|ab exwiiat' nifii'fr. r>j isfc-w aafsaaiatii rttiw— *eoa §alXqinsB gnxasoxitix iw ssod* iol nadi" eie*wo'r:, XXsn;;'. ioI bjiqmse al 10119 ©idis»»-imfcs adi iJftifX- oi ( alq.-!»sia *ia1 { JnsXq isoq avij-sUa^XXx srii .11 afif *o! nb.-!.j to Sw& iaq 6*?' ioifcf S^Uups* life io3 iisatf isq X ^ *d* 'd>l irwtf isq X Id sMina.t*-jrt3ia ^t^6 fJrta ^"uLViof II»*58 Ttoi baKjiSS ^ brij6r<5» ^i«*a »-f» lo idea -V«q d*S ^agd^^ ziti m i*i*6atkriq Jaa§*is'i a*«S asii atrls bX>7t>w isfia ; id? lo' -fXqsfes tthtfLi-kit s sXlriV /babs^gf ai^iiES 1 *t) tozi'inoG Id as-xaajr a*da srii i% **i4*e§2*&* siiS idl- $#^6* |M iXavi.t^l3t life- &*j£ttbcrKt (iM* dtHW ^ifefe «rrfi rf3r# AiiMflafc TABLE 4 The Effect of the Desired Degree of Accuracy on the Volume cf Fruit Sampled and the Cost cf Sample Grading for Typical Apple, Pear, and Clive Packing Plants Deliveries Number Total Total sample per grower of deliveries Limit of admissible error — per cent of total weight per season growers per season 5.0 3-0 2.0 1.0 0. 5 1,000 pounds^./ 1,000 pounds per cent pounds per cent pounds per cent pounds per cent pounds per cent pounds Apple or pear plant 30 50 100 150 200 400 4 8 10 6 5 2 120 400 1,000 900 1,000 800 0.4 0-3 0.2 0.1 0.1 0.1 480 1,200 2,000 900 1,000 800 1.2 0.8 0.4 0.2 0.2 0.1 1,440 3,200 4,000 1,800 2,000 800 2.6 1.6 0.8 0.5 0.4 0.2 3,120 6,4oo 8,000 4,500 4,000 1,600 9.6 6.0 3-1 2.1 1.6 0.6 11,520 24,000 31,000 18, 900 16,000 6,400 30.0 20.4 11.4 7.9 6.0 3.0 36,000 81,600 114,000 71,100 60,000 24,000 35 4,220 0.2 6,380 0.3 13,240 0.7 27,620 2.6 107,820 9.2 386,700 Total sampling cost*?/ $175 $190 $223 $4o6 $1,042 Olive p Lant 5 10 20 50 100 200 5 5 0.9 45 2.5 125 5.6 280 19.3 965 49.0 2,450 7 35 0.2 70 0.6 210 l.l 385 4.7 1,645 16.2 5,670 10 100 0.1 100 0.3 300 0.5 500 2.4 2,400 8.9 8,900 8 160 0.1 160 0.1 160 0.3 48o 1.2 1,920 4.6 7,360 5 250 0.1 250 0.1 250 0.1 250 0.5 1,250 1.9 4,750 5 500 0.1 500 0.1 500 0.1 500 0.3 1,500 1.0 5,000 1 200 0.1 200 0.1 200 0.1 200 0.1 200 0.5 1,000 41 1,250 0.1 1,325 0.1 1,745 0.2 2,595 0.8 9,880 2.8 35,130 Total sampling cost $192 $202 $222 $397 $1,003 a/ Average of three fruit per pound used to calculate number of fruit with which to enter Figure 1. b/ Costs based on larger size sample grading table. C/ Average of 40 Sevillano olives used to calculate the number of fruit with which to enter Figure 1. Field run fruit vary from 40 to 60 per pound. The lower figure is used to be more certain of limiting the possible error to the desired range. 20 accurate for large producers. From this standpoint, then, differential sampling — varying the per cent sampled according to grower size — is a more equitable procedure. Following procedures similar to those indicated in Table h, costs of sample grading have been estimated for a number of plants now using separate-lot systems. These costs are given in Table £ for several de- grees of accuracy. Details as to the season deliveries for each grower were not readily available for every plant. Therefore, the sampling costs have been estimated by using the average volume of deliveries per grower to determine the average per cent of fruit to be included in sam- ples. The cost estimates will be about the same as with the more detailed procedures. As in the typical plants represented in Table k f sampling costs per thousand pounds of fruit received increase with increases in the degree of accuracy. The costs also increase with decreases in the average volume of deliveries per grower and, due to the fixed costs for equipment, de- crease with increases in the total plant volume per season. SAMPLING VERSUS SEPARATE-LOT SYSTEMS The circumstances under which sampling or separate-lot procedures would be more economical as a means of accounting to growers for the amounts of their fruit falling in each grade or size classification cannot be described in any simple form. As has been indicated, the cost of a separate-lot system depends on such factors as the length of break periods between lots, the average size of lot, rates of plant output, and total direct hourly payroll. At the same time, sampling costs are influenced by the desired degree of accuracy, the amount of advance sampling information available, and the distribution of total deliv- eries among individual growers. The possible sets of conditions for which cost comparisons could be made are innumerable. The procedures for calculating the costs of separate-lot systems have been indicated in Tables 1 and 2. Costs of the alternative system — sample grading — may be estimated by following the procedures used in Table 1|, These calculations may be carried out by a manager for his own plant to ascertain which system is likely to be most economical. cfflT tit D30 ~}tCS lot OR! TABLE 5 Costs cf Sample Grading Systems in Relation to Degree of Accuracy for California Apple, Pear, and Olive Packing Plants Average cost of sample grading (dollars per 1,000 pounds of fruit received^' Limit of admissible error for the average-size grower — Volume per Average volume cf per cent of the total weight delivered Plant season deliveries per grower 3-0 2.0 1.0 0.5 estimated 1,000 pounds 1,000 pounds 1,000 fruit Apple and pear packing houses A 3,520 90 270 0.06 0.06 0.13 0.33 B 2,331 70 210 0.08 0.09 0.17 0.42 L 4,524 115 3^5 0.04 0.05 0.10 0.27 M 6,385 320 960 0.03 0.03 0.05 0.11 N 5,695 240 720 0.03 0.04 0.06 0.14 R 9,000 320 960 0.02 0.02 0.04 0.10 S 10,300 115 345 0.02 0.03 0.08 0.25 T 8,125 580 1,7^0 0.02 0.02 0.03 0.07 U 5,745 190 570 0.03 0.04 0.07 0.17 Olive plants I 487 35 1,400 0.35 0.40 0.47 0.95 II 1,713 42 1,680 0.12 0.12 0.21 0.62 III 3,100 18 720 0.08 0.12 O.36 1.25 rv 1,527 26 1,040 0.13 0.15 0.32 0.92 V 2,098 46 1,840 0.10 0.10 0.20 0.56 VI 1,218 Ik 560 0.18 0.23 0.54 1.67 VIII 824 16 640 0.24 0.29 0.56 1.56 IX 2,1)09 56 2,240 0.09 0.09 0.16 0.52 a/ Costs for apples and pears based on large size sample grading table. H 22. Some general conclusions are possible, however, by comparing the costs of separate-lot systems given in Table 2 with the estimated costs of sam- pling systems given in Table $ for these same plants. The comparative costs are illustrated by the bars in Figure k* For each plant, the bar on the left represents the cost of the separate-lot system. The total area of the bar indicates the present cost of the separate-lot system and the darker stippled area represents the estimated cost with standard two-minute average break time between lots. The present break time in Plants S and III was less than or equal to two minutes so no reduction is indicated. Data on costs with two-minute breaks were not available in Plants B, R, and T. The bars on the right represent the estimated cost of a sample grading system in each of these plants. The total area of the bars shows the estimated cost of sampling with an admissible error of one-half of 1 per cent of the total weight of fruit for the average-size grower. The cross-hatched area indicates the cost of sampling with a 1-per cent limit of admissible error. For apple and pear plants, the estimated cost of the sampling system with a limit of admissible error of 1 per cent for the average grower is, in all cases, less than the cost of separate-lot systems — even where the break period is reduced to two minutes. If it is desired to limit the admissible error to one-half of 1 per cent, the cost of sampling is sharply increased. However, even in this case sampling costs are less than separate- lot costs in five of the nine plants. Thus, for the majority of these plants, the sampling system tends to be the most economical procedure. The case for sampling is less clear in olive plants. While the costs of the separate-lot system per 1,000 pounds of fruit received are gener- ally higher in these plants than in apple and pear plants, so also are sampling costs. With 1-per cent accuracy, estimated sampling costs are less than present costs of separate-lot systems in seven of the eight olive plants and are less than separate-lot costs with two-minute breaks in five of the eight plants. However, if a limit of admissible error of only one-half of 1 per cent is desired, sampling costs are greater than present separate-lot costs in all but one plant. Thus, with a limit of admissible error of 1 per cent or more, the sampling system has a cost advantage in the majority of the plants. If a higher degree of accuracy is desired, either the present or the improved separate-lot system is generally less costly. ■ • • ■ ■ . ■ ■ . :■ . 1 i 1.00 to -q Jo §> 1 © | o o> || °-8 D) C C 3 — O p. a 3 o 8 o -£ Z *S a.0. «- to E O _D 0)0 a u at > < Estimated costs of the separate lot system Estimated costs of the sampling system Present Improved — break periods reduced to average of 2 minutes ' Present only — data for estimating cost with improved system not available Admissible error of 0.5 per cent of the total weight of fruit for the average-size grower Admissible error of 1 .0 per cent of the total weight of fruit for the average-size grower 50 B L M N R S — Pear and apple plants U III IV V VI VIII Olive plants IX Figure 4. Comparative costs of sampling and separate lot systems in Ca packing or processing plants. lifornia pear, apple, and olive 2U APPENDIX A The sampling chart for Case I and II sampling conditions and the rules for its use under Case III, presented in the previous pages of this report, should provide a useful guide to packing house operators who are interested in setting up sampling systems or in checking on their present procedures. For some plants, however, a more detailed and comprehensive treatment may be required or desired. This appendix is included for the benefit of those who are interested in more precise formulations of sam- pling Cases I, II, and III, or in designing a sampling system directly in terms of accuracy of the final payments to producers — Case IV. The pro- cedures for determining the quantity of fruit to be sampled from each lot for each sampling case are presented here in simple algebraic form. Cases III and IV are also treated graphically. The calculations throughout are based on the probability of exceeding the specified limit of admissible error only once out of twenty times. 1. Accuracy of an estimated proportion The general expression for determining the per cent to be sampled from each lot in order to obtain a desired limit of admissible error is c _ P (1 - P) (1) where S = the per cent to be sampled from each lot P = the expected proportion in the particular grade or class A = the limit of admissible error as a per cent of the total weight of all grades of fruit 9/ N = the expected total number of fruit delivered.-' 9/ Note that this expression may be written /P (1 - P ) ( N - n ) A = 1.96 V n N where n is the total size of sample and the expression under the square ro ot is the standard error of a proportion. When multiplied by 1.96 — i.e., V3 • 8Ul6 — the standard error includes 95 per cent of the sample estimates. Equation (1) is obtained by simply writing the above expression in a form that gives n/N, where n/N = S. As was indicated earlier, to obtain a repre- sentative sample requires that a percentage of the fruit in each lot be in- cluded in the total sample. The absolute quantity of fruit sampled, (n), increases slowly with increases in total deliveries (N). Therefore, the per cent to be sampled from each lot, (S), decreases with increases in N. • .' ' ". ; 4 ". • • • r ■ • • • ' • ' ■ . ,. . ■ ... . • ■ - S sail 25 A modification of this expression is used for each of the first three sampling cases described previously. Case I ; Nothing is known about total deliveries, expected proportions, or prices. P is assumed to be 0.5 as a sample that will limit the admis- sible error to the desired amount for this proportion will result in a smaller error for any other proportion. N is the number of fruit in the lot, which can be estimated. Expression (1) becomes S = g m2$ (2) W N + - 25 This equation was used to obtain the curves in Figure 3. Case II t Estimates are available concerning expected total deliveries per season. Expression (2) still applies but N now becomes the expected deliveries per season rather than the number of fruit per lot. Case III : Estimates are available concerning expected total deliver- ies per season and the expected quantity of fruit in the largest class or grade. Return to expression (1). Substitute for P the expected proportion in the largest grade. Use a minimum value if the expected proportion in the largest grade is more than 50 per cent, a maximum value if less than 50 per cent. For example, the expected proportion in the largest grade might be at least 70 per cent or not more than k0 per cent. The expected error in estimating any other grade will not be any greater than the error specified for the largest grade. Example ; Suppose that: (a) It is expected that a particular grower will have at least 65 per cent of his fruit in the largest grade or class. (b) Total season deliveries will be at least 90,000 fruit. (c) The desired limit of admissible error is three-fourths of 1 per cent of the total weight of fruit of all grades. .52 ©sat*, a. - ' ■ ■ tc lofts a I 26. Substituting these values in equation (1) s = «65 x .35 m .2275 ^§g|^ (90,000) + (.65 x .35) >0 ^ggf (90,000) ♦ .2275 = 5' 0625^ = *^ or ^**? per cent Thus, lli. 7 per cent of each lot of fruit delivered by this grower should be sampled. A graphic solution to the Case III sampling problem is given in Figure 5. The per cent to be sampled from each lot is determined by the expected proportion of the fruit in the largest grade, the admissible error and the total amount of fruit delivered by the grower during the season or pool period. As before, the admissible error curves in the chart are based on the probability of obtaining an error of estimate as large as, or larger than, the amount specified only once out of twenty times. To illustrate the use of the diagram, assume again the conditions of the example given above. Enter Figure 5 at point A with the proportion 0.65$ move vertically to point B on the curve representing an admissible error of 0.75 per cent; and then mcve horizontally to point C on the curve representing deliveries of 90,000 fruit. Finally, reading down to point D indicates that the desired level of accuracy may be obtained by sampling between Ik and 15 per cent of each lot of the fruit delivered. The alge- braic calculation was llu7 per cent. Note that sampling based on a maximum proportion of 50 per cent, as in the previous procedures (Figure 3), would have indicated a sample of approximately 16 per cent and that the differ- ence between the two methods increases with decreases in the admissible error, with decreases in the pool deliveries and with increases in the de- parture of the proportion in the largest grade from 50 per cent. A plant manager may also be interested in estimating the limit of error actually obtained for sample proportions after a sample has been selected. For this purpose, equation (1) may be written /P (1 - P) ,1 A = 1.96 V N K s L) (3) .b?f.qmi?a eri bii ijxrlor. •jifiqf/i'j A ■ode- nevig Percent in largest grade Percent to be sampled from each lot "Weights are for apples and pears only. In terms of numbers of fruit, the diagram has general applicability for most fruits and vegetables. Figure 5. Percent to be sampled from each lot as determined by the percent of fruit in the grade of largest quantity, the limit of admissible error, and the total season or pool deliveries per grower. 28. Example : Suppose that a sample has been drawn and : (a) The proportion in the largest grade — as estimated from the sample — is 60 per cent. (b) Total deliveries per season are 100,000 fruit. (c) Four per cent of this total was sample graded. To estimate the limit of error actually obtained: a - i qA A 6 x . It ^ 1 ~T7 -i «/ /.2U x 2h ~ 1,96 V TooTooo ( ToU " 1} = 1>96 V 100,000 1.96 100 -\^/.5>76 = .015 or 1.5 per cent In other words, the manager may assume that the true proportion of fruit in this grade is contained within the range 58.5 to 61.5 per cent (60 * 1.5), with one chance in twenty of being wrong. The historical error may also be approximated by working backwards in Figure 5» Enter the diagram with the actual per cent sampled; read vertically to the line indicating approximately the total deliveries; read horizontally to the left, stopping above the estimated proportion (as actually measured). Reading to the nearest error line will approximate the limit of error actually obtained. 2. Accuracy in terms of payments to growers — Case IV If information is available concerning the prices of fruit in the various grades or classes as well as expected deliveries and proportions, the sampling system may be designed to give, directly, some specified degree of accuracy in the payments to growers. The general expression for determining the per cent to be sampled from each lot under these conditions is (P A 2 + P 2 m/ + P 3 M 3 2 + . ..) - (P^ ♦ P 2 M 2 ♦ P 3 M 3 + ...) 2 S = — * = ^ * (it) (P,1L + PJ/L + P M, + ...) (AN ,) + (P,BL + P 0 M 0 + P 0 M^ + ...) 1 1 33 (33UIo" " 1 ) in " 2 2 3 3 where S = the per cent to be sampled from each lot P's are the proportions of fruit in each grade M's are the relative or actual prices per fruit for each grade A = the limit of admissible error as a per cent of the total payment to the grower N = the expected total deliveries (number of fruit). .Bit Isi; : yjjHI rap. ; , ■.. . 29. Note: Where accuracy is expressed as a percentage, it will be convenient to use relative rather than actual prices per fruit. Let the price of the highest grade be 1.00, with all other prices as a fraction of the highest price. It will, of course, be impossible to estimate in advance the exact proportion of fruit falling in each grade. Otherwise, there would be no need for sampling. Normally, however, the plant manager will have some idea that (say) at least 50 per cent of the fruit will fall in one grade, not more than 10 per cent in another grade, and so on. To be more certain of limiting the admissible error to the desired range, the advance esti- mates of proportions should be conservative. If there are doubts, the estimates of proportions for the highest price grade should be in the direc- tion of 50 per cent, and the estimates for other grades should be in the direction of increasing the proportions in the very lowest price grades. If the total of the two highest grades is less than 50 per cent, the esti- mates for both of these grades should be in the direction of bringing their total to 50 per cent. Example ; Suppose that in a pear packing plant it is expected that: (a) A particular grower will have at least 60 per cent of his fruit in the highest price grade and not more than 10 per cent in the lowest price grade. The other 30 per cent falls in intermediate grades. Note that these proportions are in terms of numbers of fruit rather than weight. (b) The relative prices for the three grades are: No. 1 = 1.0, No. 2 = 0.3, No. 3 = 0.1. (c) The desired limit of admissible error is 1 per cent of the total payment. (d) Expected total deliveries are 200,000 fruit. Equation (k) becomes: (.6 x l 2 + .3 x .3 2 + .1 x .l 2 ) - (.6 x 1 + .3 x .3 + .1 x .l) 2 (.6 x 1 + .3 x .3 + .1 x .1) 2 ( .0001 x 200,000 ~) (.6 x l 2 + .3 x .3 2 + .1 x .l 2 ) .628 - (-700) 2 .138 / ?nn s2 ( 2Q~~ + .628 = 20 x .1+9 + 713? = b ( * 700) (378TH6- " x) 3.8U6 Thus, to obtain the desired limit of admissible error, 5.1 per cent of each lot should be sampled. "ib ' rii Sfim 539 i. l ib s>riJ- nj T^)3t ano,Ljloqf>*iq In 8#\l'8fltx*S9 artf bus woJ n& rlcf b -'tis so en'T 30. Solutions to the Case IV sampling problem are illustrated graphically in Figures 6 and 7 for two sets of assumptions concerning the proportions of produce falling in the various grades. The limits of admissible error are again based on the probability of exceeding that amount only once in twenty times. The first graph is drawn on the assumption that all of the fruit not in the first grade has the price of the lowest grade. In this situation, if the prices of the intermediate grades are higher than that of the lowest grade, the probability of exceeding the limit of error in esti- mating the payments to growers will be even less than the amount specified. To illustrate the use of this diagram, suppose that: (a) It is estimated that a particular grower will have about 65 per cent of his fruit in the No. 1 grade. (b) The expected price of the lowest grade relative to the highest grade is 0.3. (c) Total season deliveries are expected to be at least 90,000 fruit. (d) It is desired to limit the admissible error to three-fourths of 1 per cent (0.75) or less in the total payments. Enter Figure 6 at point A with the proportion, 65 per cent (0.65) j move horizontally to point B on the curve representing a relative price of 0.3; then move down to point C on the line representing an admissible error of 0.75 per cent; from C, move across to point D on the curve repre- senting total deliveries of 90,000 fruit. Finally, reading down to point E indicates that the desired level of accuracy may be obtained by sampling 13.3 per cent of each lot of fruit delivered. In this example, if the relative prices had been 0.2. the required sample would have been 18 per cent; if 0.5, about 6 per cent. Note that it is also possible to inter- polate between the curves representing various amounts of total deliveries. In the example above, if deliveries had been 120,000 fruit (ijO,000 pounds of apples or pears) instead of 90,000, the required sample would have been about 10.5 per cent. If the price of the lowest grade of fruit (culls) is zero and the proportion of the fruit falling in this grade quite small, the per cent to be sampled from each lot as determined from Figure 6 will be larger than required to give the desired degree of accuracy. To avoid this situation, the alternative diagram (Figure 7) may be used under these conditions. This chart is based on the assumption that the proportion ■ - "is j as 50 60 ® 70 80 S. 90 e Percent to be sampled from each lot 15 © 10 5 d of ioi o'l 07 30. I0 > -1000 Fruit -1000 Lbs. 60, 20 TOTAL SEASON OR POOL PERIOD DELIVERIES PER. .GROWER 225 75 300 100 450 600 150 200 oZ © Weights are for apples and peors only. In terms\ of numbers of fruit, the diagram has general ap- plicability for most fruits and vegetables. 900 1200 300 400 1000 Fruit 1000 Lbs* 0.50 O PRICE OF LOWEST GRADE RELATIVE TO HIGHEST GRADE •2.00 LIMIT OF ADMISSIBLE ERROR PERCENT OF WEIGHTED AVERAGE PRICE 0.75 1.50 1.25 .00 Figure 6. Percent to be sampled from each lot as determined by the percent of fruit in the No. 1 grade, price of the lowest grade relative to No. 1, the limit of admissible error and the total pool deliveries per grower. Figure 7. Percent to be sampled from each lot as determined by the percent of fruit in the No. 1 grade, pi of the next to lowest grade relative to No. I, the limit of admissible error and total season de- liveries per grower. The quantity in the lowest grade (culls) is not more than 10 percent and the price is zero. 33. of fruit falling in the lowest grade will be no larger than 10 per cent and that the price for this grade will be zero. The fruit not falling in the first grade is assumed to have a price equal to the next to the lowest grade. Again, if the prices of the intermediate grades are greater than that of the next to the lowest, the probability of exceeding the limit of error will be less than the amount specified. Each curve in the upper right-hand section of the chart here represents the price of the next to the lowest grade of fruit relative to the highest grade rather than the lowest grade relative to the highest. The diagram is similar to and used in the same way as Figure 6. As before, this equation (U) can be turned around and used to estimate the limit of error for a set of sample estimates already obtained. For this purpose, equation (k) may be written: A * l*96yVfr^f + P 2 M 2 2 + ...) - (P^ + P 2 M 2 + ...)f7(| - 1) (5) N Example ; Suppose that a sample has been taken and: (a) The estimated proportions for four grades of fruit as determined by the sample are No. 1 = .0$, No. 2 = 0.3, No. 3 = 0.1, and No. k = 0.1. (b) The prices for the various grades are No. 1 = 1.0, No. 2 = 0.6, No. 3 = 0.3, and No. h = 0.1. (c) A 5-per cent sample was taken. (d) Total deliveries for this grower were 100,000 fruit. (e) The desired limit of admissible error is 1 per cent of the total payment. To estimate the limit of error, equation ($) becomes K*\.%x/$^**- + -l*-^ + + .lx.l 2 )-(.$xl + .3x.6 + .lx.3 + .lx.l)f7(-^ - 1) N =1 • 96 -^T.618)- (.?20)f7 (19) 100,000 ~i,ooo, V 19 = ,009 or ' 9 per cent Thus, the estimated limit of error is nine-tenths of 1 per cent of the total payment. Previous Reports in This Series on EFFICIENCY IN FRUIT MARKETING Marketing Costs for Deciduous Fruits Grading Costs for Apples and Pears Orchard-to-Plant Transportation Packing Costs for California Apples and Pears Building and Equipment Costs, Apple and Pear Packing In-Plant Transportation Costs as Related to Materials Handling Methods — Apple and Pear Packing ' ■