If you must use a BED file and intersectBed:

 intersectBed -a ES.bed -b Ensembl_mm9.bed -wb -s | awk '{a[$10]++}END{for(idx in a) {print idx,a[idx]}}'

Keep in mind that intersectBed shouldn't be used to count RNAseq reads, since it will increment the counter for a feature even if a read maps to not just it but another feature. As an example:

>cat foo.bed
chr1    0    100    read1    .    +
chr1    0    100    read2    .    -
chr1    50    150    read3    .    +
chr1    100    200    read4    .    +

>cat bar.bed
chr1    0    100    target1    .    +
chr1    0    100    target2    .    -
chr1    20    120    target3    .    +

>intersectBed -a foo.bed -b bar.bed -wb -s | awk '{a[$10]++}END{for(idx in a) {print idx,a[idx]}}'
target1 2
target2 1
target3 3

Only target2 should have a count (or target1 should have 1 and target3 should have 2, depending on how you overlap things). You could script around this, but it's faster to just use a different tool.

Edit: I can recommend featureCounts (from Subread) and also htseq-count for this. Making a GTF should be relatively straight-forward. Just increment the second column, set the 4th as the gene_id and transcript_id and shuffle around the remainder (adding some "." columns).