I had a combination of issues in the data step that created the cntlin datasets.  The code below solved my problems. proc sql; drop table WORK.changethese3; quit; DATA work.CHANGETHESE3(KEEP=FMTNAME START END LABEL min max hlo) ;     LENGTH FMTNAME $ 8 START 5 LABEL 8; retain type "N" FMTNAME "chgid"  min 3 max 8;     set work.changethese2 END=LASTREC; if _n_ = 1 then do;      START=0;      END=(ID-1);         LABEL=0;   OUTPUT; end;     START=ID; END=ID;     LABEL=1;     OUTPUT;      IF LASTREC THEN DO;         LABEL=0;      START=(ID + 1); END=999;      hlo = '0';         OUTPUT;    END; run; PROC FORMAT CNTLIN=WORK.changethese3; QUIT; proc format fmtlib; select chgid; QUIT; proc sql; drop table WORK.changethese4; quit; DATA work.CHANGETHESE4(KEEP=FMTNAME START END LABEL hlo min max OTHER) ;     LENGTH FMTNAME $ 8 START 5 LABEL $1;     set work.changethese2 END=LASTREC;      retain FMTNAME "chgrsk" min 1 max 5;    OTHER = ' ';     START = ID; END = ID;     LABEL = riska;     OUTPUT;     IF LASTREC THEN DO;       START=.;    END = .;       LABEL=' ';    hlo = '0';       OUTPUT;       END; run; PROC FORMAT CNTLIN=WORK.changethese4; run; proc format fmtlib; select chgrsk; run;              1596  data work.world;    /* unprojected */ 1597         length chekid 8 chekrisk $1; 1598         format chekid 5. chekrisk $1. chekid2 chgid.; 1599     set work.worldx; 1600    chekid = put(id,chgid.); 1601    chekid2 = id; 1602     if chekid = 1 then do; 1603                chekrisk = put(id,chgrsk.); 1604               riska=put(id,chgrsk.); 1605               put id= riska=; 1606   end; 1607    if _n_ < 5 then do;   put "first four rows" id= chekid= chekid2=  chekrisk =; end; 1608    drop chekid chekrisk chekid2; 1609  run; NOTE: Character values have been converted to numeric values at the places given by:       (Line):(Column).       1600:12 first four rowsID=72 chekid=0 chekid2=0 chekrisk= first four rowsID=72 chekid=0 chekid2=0 chekrisk= first four rowsID=72 chekid=0 chekid2=0 chekrisk= first four rowsID=72 chekid=0 chekid2=0 chekrisk= ID=399 riska=2 ID=573 riska=1 ID=573 riska=1 ID=573 riska=1 NOTE: There were 45805 observations read from the data set WORK.WORLDX. NOTE: The data set WORK.WORLD has 45805 observations and 9 variables. ... View more

Using libname with the EXCEL engine solved the problem but I still do not know the reason why Proc Export causes the problem. libname mylib excel "e:\testing\trash\MyFile testing output.xlsx";   proc sql; drop table mylib.testing; quit;   data mylib.testing;         set anova_full;   run; libname mylib clear; ... View more